raman spectroscopy based diagnosis of dengue virus
Post on 18-Jan-2022
4 Views
Preview:
TRANSCRIPT
Bilal Khan
2018
Department of Physics and Applied Mathematics
Pakistan Institute of Engineering and Applied Sciences
Nilore, Islamabad, Pakistan
Raman Spectroscopy based Diagnosis of
Dengue Virus Infection in Human Blood
Serum
This page intentionally left blank.
Reviewers and Examiners
Foreign Reviewers
1. Dr. Georges Wagnieres,
Head Photo-medicine Group, Laboratoire Leenaards-Jeantet d’imagerie fonctionnelle
et métabolique, Ecole Polytechnique Fédérale de Lausanne, Switzerland.
2. Mark Cronin-Golomb, Associate Professor
Department of Biomedical Engineering, Science and Technology Center,
Massachusetts, USA.
3. Carlos R. Stroud, Professor of Optics & Professor of Physics
The Institute of Optics, University of Rochester, Rochester, New York, USA.
Thesis Examiners
1. Dr. Muhammad Aslam Baig, Professor
National Center for Physics (NCP), Islamabad.
2. Dr. Farhan Saif, Professor
Department of Electronics, Quaid-i-Azam University, Islamabad.
3. Dr. Shahid Manzoor, CSO
Physics Department, CIIT, Islamabad.
Head of the Department (Name): Dr. Muhammad Yousaf Hamza
Signature with Date: _________________________________
Certificate of Approval
This is to certify that research work presented in this thesis titled “Raman Spectroscopy
based Diagnosis of Dengue Virus Infection in Human Blood Serum” was conducted by
Mr. Bilal Khan under the supervision of Dr. Mushtaq Ahmed.
No part of this thesis has been submitted anywhere else for any other degree. This thesis is
submitted to Department of Physics and Applied Mathematics in partial fulfillment of the
requirements for the degree of Doctor of Philosophy in the field of Physics.
Student Name: Bilal Khan Signature: ----------------------------
Examination Committee:
Examiners Name, Designation & Address Signature
Internal Examiner 1 Dr. Muhammad Aslam Baig
Internal Examiner 2 Dr. Farhan Saif
Internal Examiner 3 Dr. Shahid Manzoor
Supervisor Dr. Mushtaq Ahmed
Co-Supervisor Dr. Masroor Ikram
Department Head Dr. Muhammad Yousaf Hamza
Dean Research PIEAS Dr. Mutawarra Hussain
Thesis Submission Approval
This is to certify that the work contained in this thesis entitled Raman Spectroscopy based
Diagnosis of Dengue Virus Infection in Human Blood Serum, was carried out by Bilal
Khan, and in my opinion, it is fully adequate, in scope and quality, for the degree of Ph.D.
Furthermore, it is hereby approved for submission for review and thesis defense.
Supervisor: _____________________
Name: Dr. Mushtaq Ahmed
Date: April 02, 2018
Place: PIEAS, Islamabad.
Co-Supervisor: __________________
Name: Dr. Masroor Ikram
Date: April 02, 2018
Place: PIEAS, Islamabad.
Head, Department of Physics and Applied Mathematics: ___________________
Name: Dr. Muhammad Yousaf Hamza
Date: April 02, 2018
Place: PIEAS, Islamabad.
Raman Spectroscopy based Diagnosis of
Dengue Virus Infection in Human Blood
Serum
Bilal Khan
Submitted in partial fulfillment of the requirements
for the degree of Ph.D.
2018
Department of Physics and Applied Mathematics
Pakistan Institute of Engineering and Applied Sciences
Nilore, Islamabad, Pakistan
ii
Dedications
To my loving mother, visionary father, caring mother-in-law,
encouraging wife
&
four loving daughters; Aayzah, Sonaiha, Maysha and Ainasaba.
iii
Author’s Declaration
I Bilal Khan hereby declare that my PhD thesis titled “Raman Spectroscopy based
Diagnosis of Dengue Virus Infection in Human Blood Serum“ is my own work and has
not been submitted previously by me or anybody else for taking any degree from Pakistan
Institute of Engineering and Applied Sciences (PIEAS) or any other university / institute in
the country / world.
At any time if my statement is found to be incorrect (even after my graduation), the
university has the right to withdraw my PhD degree.
__________________
(Bilal Khan)
April 02, 2018
PIEAS, Islamabad.
iv
Plagiarism Undertaking
I Bilal Khan solemnly declare that research work presented in the thesis titled “Raman
Spectroscopy based Diagnosis of Dengue Virus Infection in Human Blood Serum” is
solely my research work with no significant contribution from any other person. Small
contribution / help wherever taken has been duly acknowledged or referred and that complete
thesis has been written by me.
I understand the zero tolerance policy of the HEC and Pakistan Institute of
Engineering and Applied Sciences (PIEAS) towards plagiarism. Therefore, I as an author of
the thesis titled above declare that no portion of my thesis has been plagiarized and any
material used as reference is properly referred/cited.
I undertake that if I am found guilty of any formal plagiarism in the thesis titled above
even after the award of my PhD degree, PIEAS reserves the rights to withdraw/revoke my
PhD degree and that HEC and PIEAS has the right to publish my name on the HEC/PIEAS
Website on which name of students are placed who submitted plagiarized thesis.
__________________
(Bilal Khan)
April 02, 2018
PIEAS, Islamabad.
v
Copyrights Statement
The entire contents of this thesis entitled Raman Spectroscopy based Diagnosis of Dengue
Virus Infection in Human Blood Serum by Bilal Khan are an intellectual property of
Pakistan Institute of Engineering & Applied Sciences (PIEAS). No portion of the thesis
should be reproduced without obtaining explicit permission from PIEAS.
vi
Table of Contents
Dedications............................................................................................................................ ii
Author’s Declaration .......................................................................................................... iii
Plagiarism Undertaking ..................................................................................................... iv
Copyrights Statement .......................................................................................................... v
Table of Contents ................................................................................................................ vi
List of Figures ...................................................................................................................... ix
List of Tables ....................................................................................................................... xi
Abstract ............................................................................................................................... xii
List of Publications ........................................................................................................... xiv
List of Abbreviations and Symbols .................................................................................. xv
Acknowledgments ............................................................................................................ xvii
1 Introduction ................................................................................................................ 1
1.1 Overview .............................................................................................................. 1
1.2 Objective .............................................................................................................. 3
1.3 Chemical Diagnostic Techniques ........................................................................ 4
1.4 Optical Diagnostic Techniques ............................................................................ 4
1.4.1 Elastic Scattering Spectroscopy (ESS) ......................................................... 4
1.4.2 Diffuse Reflectance Spectroscopy (DRS) .................................................... 4
1.4.3 Differential Path-Length Spectroscopy (DPS) ............................................. 5
1.4.4 Near Infrared Spectroscopy .......................................................................... 5
1.4.5 Fluorescence Spectroscopy ........................................................................... 5
1.4.6 Raman Spectroscopy .................................................................................... 6
1.5 Thesis Layout ....................................................................................................... 6
2 Diagnosis of DENV Infection and Raman Spectroscopy ........................................ 7
vii
2.1 DENV Infection ................................................................................................... 7
2.1.1 DENV Structure ........................................................................................... 7
2.1.2 DENV Strains and ‘Immune Protection’ ...................................................... 7
2.1.3 Diagnostics Techniques ................................................................................ 8
2.1.4 Dengue in the Future .................................................................................. 11
2.2 Raman Spectroscopy .......................................................................................... 11
2.2.1 History of Raman Scattering ...................................................................... 11
2.2.2 Raman Shift ................................................................................................ 12
2.2.3 Classical Description .................................................................................. 12
2.2.4 Quantum Description .................................................................................. 16
2.2.5 Raman Signal Intensity ............................................................................... 18
2.2.6 Quantitative and Qualitative Analysis ........................................................ 19
2.2.7 Recent Advances in Raman Spectroscopy ................................................. 20
2.2.8 Diagnostic Applications ............................................................................. 21
2.2.9 Advantages ................................................................................................. 22
2.2.10 Disadvantages............................................................................................ 23
2.3 Summary ............................................................................................................ 23
3 Materials and Methods ............................................................................................ 25
3.1 Collection of Samples ........................................................................................ 25
3.2 Experimental Setup ............................................................................................ 25
3.3 Preprocessing Methods ...................................................................................... 27
3.4 Statistical Analysis ............................................................................................. 29
3.4.1 Multivariate Analysis Techniques .............................................................. 29
3.4.2 PLS Regression........................................................................................... 29
3.5 Molecular Analysis ............................................................................................ 33
4 Results and Discussions ............................................................................................ 34
4.1 NS1 based Screening ......................................................................................... 34
viii
4.1.1 Results and Discussion: .............................................................................. 34
4.2 IgM based Screening .......................................................................................... 40
4.2.1 Results and Discussions: ............................................................................ 40
4.3 IgG based Screening .......................................................................................... 47
4.3.1 Results and Discussions ............................................................................. 47
4.4 Lactate as Biomarker ......................................................................................... 57
4.4.1 Acquiring Raman Spectra ........................................................................... 57
4.4.2 Raman spectral analysis.............................................................................. 58
4.4.3 Results and Discussion ............................................................................... 60
5 Conclusions and Future Prospects .......................................................................... 63
5.1 Raman Spectroscopy based Diagnosis of DENV Infection ............................... 64
5.1.1 NS1 based Study ......................................................................................... 64
5.1.2 IgM based Study ......................................................................................... 64
5.1.3 IgG based Study ......................................................................................... 65
5.1.4 Lactate as a Biomarker ............................................................................... 65
5.2 Future Prospective ............................................................................................. 65
References ........................................................................................................................... 67
ix
List of Figures
Figure 2-1 A typical Raman spectrum. ................................................................................... 12
Figure 2-2 Diatomic molecule as a mass on a spring. ............................................................ 13
Figure 2-3 Bond length of a diatomic molecule during a vibration. ....................................... 15
Figure 2-4 Polarizability as a function of vibrational displacement about equilibrium. ......... 15
Figure 2-5 Energy level diagram. ........................................................................................... 16
Figure 2-6 Conservation of energy for Raman scattering (Stokes). ....................................... 17
Figure 2-7 Energy level diagram for Raman scattering. ......................................................... 18
Figure 3-1 Schematic diagram of a typical fiber optic probe based Raman system. .............. 26
Figure 3-2 Experimental setup of the Raman system. ............................................................ 27
Figure 3-3 Preprocessing of the Raman spectrum from a raw spectrum to final a
preprocessed Raman spectrum is shown here: a: Raw Raman spectrum, b: denoised and
smoothed Raman spectrum, c: baseline corrected Raman spectrum, d: vector-normalized
Raman spectrum....................................................................................................................... 29
Figure 3-4 Mathematical description of PLS regression. ....................................................... 30
Figure 3-5 Area under ROC curve and its interpretation. ....................................................... 32
Figure 4-1 Raman spectra of sera samples used in NS1 based screening study. .................... 35
Figure 4-2 Calibration curve of model for NS1 based screening. .......................................... 36
Figure 4-3 Regression coefficients of PLS model for NS based screening. ........................... 36
Figure 4-4 Raman spectra of sera samples used in IgM based screening study. .................... 40
Figure 4-5 RMSECV curve calculated by using number of PCs from 1 to 20. ...................... 41
x
Figure 4-6 Curves obtained by employing methods of Kaiser, Scree and parallel factor
analysis for IgM based screening. ............................................................................................ 41
Figure 4-7 Calibration curve for IgM based screening along with predictions for testing data
set. ............................................................................................................................................ 42
Figure 4-8 Sensitivity, specificity and accuracy of the IgM based screening model at
different cut-off values. ............................................................................................................ 43
Figure 4-9 ROC curve for IgM based screening. .................................................................... 43
Figure 4-10 Regression vector along with average spectra of negative, mild IgM positive and
strong IgM positive samples. ................................................................................................... 44
Figure 4-11 Patch area display of Raman spectra used in IgG based screening. .................... 48
Figure 4-12 RMSE curve for PCs optimization for IgG based model. ................................... 48
Figure 4-13 Eigen values based curves for optimization for IgG based model. ..................... 49
Figure 4-14 Calibration curve of PLS model developed for IgG based screening. ................ 50
Figure 4-15 Sensitivity, specificity and accuracy of the PLS model for IgG at different cut-
off values. ................................................................................................................................. 50
Figure 4-16 Receiver operator characteristic (ROC) curve for IgG based screening model. . 51
Figure 4-17 Regression vector along with average spectra of negative, mild IgG positive and
strong IgG positive samples. .................................................................................................... 52
Figure 4-18 Sketch of experiment setup. ................................................................................ 58
Figure 4-19 Vector normalized mean Raman spectra of healthy and dengue infected sera
(upper) along with the mean difference between the normal and infected samples (lower). .. 58
Figure 4-20 Vector normalized Raman spectra of lactic acid solution. .................................. 59
Figure 4-21 Vector normalized mean Raman spectra of healthy sera, dengue infected sera, 50
mM/L and 100 mM/L of lactic acid solution in healthy sera. ................................................. 60
xi
List of Tables
Table 2-1 Cross-sections of most common optical processes [87]. ........................................ 19
Table 2-2 σ(νex) of Raman scattering for CHCl3 at different incident wavelengths [87]. ....... 19
Table 4-1 Positive regression coefficient obtained by PLS model for NS1 based screening
along with molecular description. ............................................................................................ 38
Table 4-2 Negative regression coefficient obtained by PLS model for NS1 based screening
along with molecular description. ............................................................................................ 39
Table 4-3 Prominent Raman bands highlighted by the strongly positive or strongly negative
values of regression coefficients of IgM based model. ............................................................ 46
Table 4-4 Prominent Raman bands which have been highlighted by the strongly negative
values of regression coefficients of this model are tabulated for their bio-molecular
assignment................................................................................................................................ 54
Table 4-5 Prominent Raman bands which have been highlighted by the strongly positive
values of regression coefficients of this model are tabulated for their bio-molecular
assignment................................................................................................................................ 55
xii
Abstract
Dengue virus (DENV) infection is a mosquito born infectious disease. Its diagnostic is utmost
important for treatment, as the symptoms of disease are quite similar to other diseases.
Current pathological diagnostics methods available are reverse transcriptase polymerase
chain reaction (RT-PCR) and enzyme linked immunosorbent assay (ELISA). RT-PCR is
used to detect the virus itself while ELISA is used to detect non-structural protein-1 (NS1)
and antibodies like immunoglobulin-M (IgM) and immunoglobulin-G (IgG). Existing
methods e.g. virus isolation, RT-PCR and ELISA have certain disadvantages like more time
consuming, false-positive/false-negative results and expensive as compared to Raman
spectroscopy. Raman spectroscopic technique provides molecular signatures, minimum
running cost and online results. Raman spectra of biological samples combined with a
suitable statistical data-mining technique like partial least squares (PLS) regression can be
used to devise a new method for diagnosis of DENV infection in human blood sera. In
present studies, this technique is successfully applied for the diagnostic of DENV infection
based on three steps. A graphical user interface (GUI) was specially designed and its code
was developed in MATLAB (Mathworks 2009a) programming language to implement PLS
for the presented research work.
First step: Raman spectra of ELISA confirmed NS1 positive and negative sera samples
are discriminated by PLS regression. Analysis of regression coefficients, which differentiate
these groups, shows an increasing trend for phosphatidylinositol, ceramide and amide-III, and
a decreasing trend for thiocyanate in the DENV infected serum.
Second step: Raman spectra of samples, with known value of ELISA based AI of IgM
are discriminated by PLS regression. Analysis of regression coefficients revealed that
concentration of asparagine, glutamate, galactosamine etc. were found to increase while
concentration of fructose, cholesterol, cellobiose, and arabinose were found to decrease with
increasing values of antibody index (AI) of IgM.
Third step: Raman spectra of samples, with known value of ELISA based AI of IgG are
discriminated by PLS regression. Analysis of regression coefficients revealed that myristic
acid, coenzyme-A, alanine etc. were found to increase, while amide III, collagen, proteins,
xiii
fatty acids, phospholipids and fucose were found to decrease with increasing values of AI of
IgG. Raman spectroscopy provides not only the diagnosis of DENV infection, but it also
enables the detailed insight of the abnormalities appearing in molecular composition of a
sample.
Importantly, Raman spectra @ 532 nm excitation were used to investigate the possible
use of lactate as biomarker for DENV infection. It was found that spectral difference in
healthy and infected samples is due to an elevated level of lactate in DENV infected group.
xiv
List of Publications
Journal Publications
M. Bilal, M. Saleem, M. Bilal, M. Khurram, and S. Khan, “Raman spectroscopy
based discrimination of NS1 positive and negative dengue virus infected serum,”
Laser Phys. Lett., vol. 13, no. 9, p. 95603, (2016).
M. Bilal, M Saleem, M. Bilal, T Ijaz, S. Khan, R. Ullah, A. Raza, M. Khurram, W.
Akram, and M. Ahmed, “Raman spectroscopy-based screening of IgM positive and
negative sera for dengue virus infection,” Laser Phys., vol. 26, no. 11, p. 115602,
(2016).
M. Bilal, M. Saleem, M. Bial, S. Khan, R. Ullah, H. Ali, M. Ahmed, and M. Ikram,
“Raman spectroscopy based screening of IgG positive and negative sera for dengue
virus infection,” Laser Phys. Lett., vol. 14, no. 11, p. 115601, (2017).
M. Bilal, R. Ullah, S. Khan, H. Ali, M. Saleem, and M. Ahmed, “Lactate based
optical screening of dengue virus infection in human sera using Raman spectroscopy,”
Biomed. Opt. Express, vol. 8, no. 2, p. 1250, (2017).
M Bilal, M. Saleem, S. T. Amanat, H. A. Shakoor, R. Rashid, A. Mahmood and M.
Ahmed, “Optical diagnosis of malaria infection in human plasma using Raman
spectroscopy,” J. Biomed. Opt., vol. 20, no. 1, p. 17002, (2015).
S. Khan, R. Ullah, A. Khan, A. Sohail, N. Wahab, M. Bilal, M. Ahmed, “Random
Forest-Based Evaluation of Raman Spectroscopy for Dengue Fever Analysis,” Appl.
Spectrosc., vol. 71, no. 9, p. 2111-2117, (2017).
S. Khan, R. Ullah, M. Saleem, M. Bilal, R. Rashid, I. Khan, A. Mahmood and M.
Nawaz, “Raman spectroscopic analysis of dengue virus infection in human blood
sera,” Opt. - Int. J. Light Electron Opt., vol. 127, no. 4, pp. 2086–2088, (2016).
S. Khan, R. Ullah, A. Khan, N. Wahab, M. Bilal, and M. Ahmed, “Analysis of
dengue infection based on Raman spectroscopy and support vector machine (SVM),”
Biomed. Opt. Express, vol. 7, no. 6, p. 2249, (2016).
xv
List of Abbreviations and Symbols
ADE Antibody-dependent enhancement
AI Antibody/Antigen index
ANN Artificial neural networks
AUC Area under ROC curve
BRC Breast cancer
CARS Coherent anti-Stokes Raman spectroscopy
CCD Charged coupled devices
DENV Dengue virus
DF Dengue fever
DHF Dengue hemorrhagic fever
DPS Differential path-length spectroscopy
DSS Dengue shock syndrome
ELISA Enzyme linked immunosorbent assay
EM Electromagnetic
ESS Elastic scattering spectroscopy
GUI Graphical user interface
IgG Immunoglobulin-G
IgM Immunoglobulin-M
IR Infra-red
LOC Lab-on-chip
LOO Leave one out
NIH National institute of health
NILOP National Institute of Lasers and Optronics
NIR Near infra-red
NORI Nuclear medicine oncology & radiotherapy institute
NS1 Non-structural protein-1
PCA Principal component analysis
PCs Principal components
PLS Partial least squares
xvi
PRS Polarized Raman spectroscopy
QCMD Quality control for molecular diagnostics
RF Random forest
RMC Rawalpindi medical college
RMSECV Root mean squared error in cross validation
RMSEP Root mean squared error in predictions
RNA Ribonucleic acid
ROC Receiver operating characteristic
RRS Resonance Raman spectroscopy
RT-PCR Reverse transcriptase polymerase chain reaction
SD Standard deviation
SERS Surface enhanced Raman spectroscopy
SVM Support vector machine
TEC Thermo electric cooler
xvii
Acknowledgments
First and foremost, I am heartedly thankful to Almighty Allah (جل جلاله) for His countless
blessings including the opportunity of the present research work and its completion. I offer
my sincere gratitude to the Holy Prophet Hazrat Muhammad (صلى الله عليه وسلم), the sole pride of mankind
and icon of mercy.
I would like to pay heartfelt gratitude to my supervisor Dr. Mushtaq Ahmed and co-
supervisor Dr. Masroor Ikram and Dr. Muhammad Saleem. Unequivocally, their dedication
and devotion to work has been a source of immense inspiration for me. I must appreciate
their guidance, cooperation and generosity for study and research related issues and beyond.
I would like to thank specially to Dr. Muhammad Saleem (NILOP) for his kind
supervision and guidance from very first day in every stage of the present research work. He
was the driving force for me and kept me on track by his vision and support. Most
importantly, I am cordially thankful to my visionary father and caring mother who always
prayed and scarified to flourish my career. I also appreciate the exceptionally supporting role
of my wife, mother-in-law and siblings during my studies. I am thankful to Mr. Rub Nawaz
Khan for all his support, encouragement and prayers at crucial stages of my life.
I thank Dr. Rahat Ullah for his extra ordinary support in the completion of the present
research work along with Dr. Saranjam Khan, Dr. Hina Ali. I would like to thank Dr.
Muhammad Khurram and Dr. Faiza for providing biological samples and expert serology
opinion. I am thankful to my friends Dr. Banat Gul and Mr. Abdul Basit for their guidance,
advices, and valuable suggestions and comments. I am thankful to all the teachers and staff of
PIEAS, NILOP and RMC for helping me in my study and experimental work. I am also
thankful to Mrs. Uzma Shazia, Mr Iqbal, Mrs. Fatima Batool and Mr. Muhammad Irfan.
Finally, financial support from Higher Education Commission of Pakistan under PhD
5000 fellowship program (Phase-II) is highly acknowledged.
1
1 Introduction
1.1 Overview
Dengue virus (DENV) infection is a mosquito born infectious disease. Dengue fever
(DF) is the most significant arborviral disease in the world today, which is caused by
DENV. Dengue fever may develop into dengue hemorrhagic fever (DHF) and dengue
shock syndrome (DSS) in some severe cases. This virus belongs to the family
flaviviridae, and it is based on a capsulated single strand of ribonucleic acid (RNA).
There are four serotypes of this virus. Its RNA contains the coding for seven non-
structural and three structural proteins [1]. Dengue virus infected subjects are reported
with various symptoms which are somehow similar to other diseases like flu, malaria,
chickengunya and typhoid. According to an estimate by world health organization
(WHO), there are 390 million cases of DENV infections per year. Among these,
500000 cases are about hospitalizations while 25000 cases of death are estimated. It is
reported that a majority of these infections (70–80 %) are of asymptotic nature [2]–
[4]. Its vector is found throughout the subtropical and tropical areas of the world.
Female aedes aegypti mosquito is responsible for its spreading because it carries
the DENV in its mid-gut for about 7-14 days for incubation. Later it moves towards
salivary glands of the mosquito and it is mixed with saliva. Dengue virus is
transmitted to the host when its carrier mosquito bites the host i.e. human. The blood
stream, epithelial tissues and dendritic cells are infected initially at the site of biting.
This virus moves into the bone marrow, liver and other parts of the body for its
replication in the first few days [5], [6]. It produces structural and non-structural
proteins for its fast replication and facilitation respectively. Mild fever starts after 4-7
days of mosquito bite which is known as DF. Later on, the immune system of a
human body responds to DENV infection by producing anti-dengue antibodies
immunoglobulin-M (IgM) and Immunoglobulin-G (IgG). IgM is produced after 5-6
days of fever and IgG is produced after 7-10 days of fever [7]–[9]. In the first five
days of fever, non-structural protein (NS1) is the only biomarker which can be used
Introduction
2
for diagnosis of DENV infection when there is no antibody produced. Later on IgM
and IgG can be detected for diagnosis of DENV infection.
Virus isolation [10]–[13] and reverse RT-PCR [14]–[16] are costly, time
consuming and highly sophisticated methods of diagnosis which cannot be used for
screening of DENV infection as a routine test. However, ELISA is a technique which
is being commonly used for the diagnosis of DENV infection (a detailed description
is given in section 2.1.3). It screens the serum for positive and negative samples on
the basis of NS1, IgM and IgG. It is primarily based on chemical complexes like
enzyme-antigen-complex and enzyme-antibody-complex. It yields a numeric value
which represents the antigen/antibody index (AI). A sample with the value of AI
above cut-off value of that specific ELISA kit is declared as positive for that specific
antigen/antibody [17]. It is evident that impurity in reagents, used in ELISA, may
result in false-positive or false-negative results. This invites the researchers to
investigate other techniques which should be more economical, quick, reliable and
robust.
Light based diagnostic methods, known as optical diagnosis, are being
investigated in various research centers for the diagnosis of various diseases. Light
based spectroscopic techniques have the potential to analyze biological samples in a
variety of ways to characterize them for the diagnosis of diseases. These
spectroscopic techniques have been described in section 1.4 along with their potential
for diagnostic applications. These techniques include infra-red (IR) spectroscopy,
fluorescence spectroscopy, elastic scattering spectroscopy (ESS), near infra-red (NIR)
spectroscopy, differential path-length spectroscopy (DPS) and Raman spectroscopy. It
is important to mention that in resonance phenomenon based techniques tunable
wavelengths are required for different molecules, however in Raman spectroscopy a
non-resonant Raman signal can be recorded with a single wavelength from different
molecules present in the sample at a time. Raman spectroscopy has certain other
advantages of sensitivity, specificity, quickness, no-sample preparation and cost
effectiveness, which encouraged the investigation of this technique to be used as a
possibly new diagnostic tool for DENV infection.
Dengue virus infection alters the biochemistry of infected blood; therefore,
Raman spectroscopy can be used for assessment of these biochemical changes
Introduction
3
effectively. Raman spectra of human blood sera having disease e.g. malaria [18],
dengue [18]–[23], Hepatitis-C [24], diabetes, female breast cancer, nasopharyngeal
cancer and other types of cancers [25]–[27] have been investigated with the help of
statistical models to predict diseases. With the help of such type of models, partial
least squares (PLS) regression algorithm has been employed to quantify various
analytes in body fluids [28], [29]. Raman spectroscopy in combination with
multivariate analysis can be an excellent tool for the diagnosis of DENV infection
which is fast, reliable, accurate, more efficient and economical. In this contribution,
Raman spectra of positive and control sera, on the basis of NS1, IgM and IgG, were
used with a PLS regression routine to develop multivariate models in MATLAB
(Mathworks 2009a) environment for the prediction of infected samples. The model
yields a vector of regression coefficients at corresponding Raman shifts. These
regression coefficients can be analyzed to identify the biological molecules which
play certain roles in this disease.
The present research work is intended to investigate the prospects of Raman
spectroscopy with 785 nm and 532 nm excitation wavelengths for the diagnosis of
DENV infection in human blood serum. This study comprised of four main goals
which are; NS1 based screening, IgM based screening, IgG based screening and
investigation of the role of lactate as a new biomarker for DENV infection. The
results of all these studies are found to be very encouraging for further studies and
implementation of Raman spectroscopy based diagnosis of DENV infection.
1.2 Objective
DENV produces certain types of biological molecules for its reproduction and
propagation. The molecule known as NS1 helps in the propagation of DENV.
Immune system of human body protects it by producing antibodies like IgM and IgG
against pathogen i.e. DENV. Diagnosis of DENV infection is usually done by ELISA
of IgM, IgG and NS1. It is a chemical test with certain disadvantages of cost, time and
accuracy. In presented study, Raman spectra of samples with known ELISA based
results about NS1, IgM and IgG were used to develop PLS models. These models
were aimed at screening of the positive and negative sera samples by using their
Raman spectra for NS1, IgM and IgG. Results of these Raman spectroscopy based
screenings combined with clinical symptoms will help the physicians to diagnose the
Introduction
4
DENV infection in a comparatively quick, cost-effective and accurate manner.
Lactate is a potential biomarker for diagnosis of DENV infection. Elevated levels of
lactate in human blood serum can also be examined by using Raman spectroscopy. It
will further strengthen the diagnosis of DENV infection.
Every chemical compound has its own characteristic Raman spectrum which
works as its fingerprint. High level of specificity of Raman spectroscopy is capable of
increasing the accuracy, specificity and sensitivity of diagnosis of DENV infection
which is demonstrated successfully in the presented research work. Raman
spectroscopy doesn’t require any sample preparation and it does not require any
chemical, thus the proposed technique is quick, cost effective and accurate.
1.3 Chemical Diagnostic Techniques
Virus isolation, RT-PCR and ELISA are the chemical methods which are in practice
for the diagnosis of DENV infection. In virus detection [10]–[13], the DENV itself is
separated from the sample of infected subject and it takes almost a week in this
process. RT-PCR [14]–[16] is used to determine the viral load quantitatively by a
repeated process and it usually takes one day in this process. Enzyme linked
immunosorbent assay (ELISA) [30], [31] based detection of NS1, IgM and IgG are
the most commonly used techniques at the moment.
1.4 Optical Diagnostic Techniques
1.4.1 Elastic Scattering Spectroscopy (ESS)
When a tissue is exposed to light, the light is reflected, scattered or absorbed. These
scattering events may take place for a number of times in an elastic manner such that
the light coming out of the tissue has the same energy as that of the incident one.
Scattering centers can either be normal or pathological. All the scattered photons are
collected and recorded as a spectrum. By using the refractive indices of cellular
components, the changes can be identified [32]–[34].
1.4.2 Diffuse Reflectance Spectroscopy (DRS)
In this technique the same wavelength of light has been observed after scattering
from the sample [35]. It is same case of Rayleigh scattering. Scattering strength
Introduction
5
depends upon the refractive index of the sample. By using refractive index of the
cellular matrix it can determine the size and density of the matrix. Cancer is a disease
which causes these types of changes and DRS can be used for its diagnosis [36].
1.4.3 Differential Path-Length Spectroscopy (DPS)
DPS is an optical technique which is used to determine intrinsic optical properties in-
vivo in a minimally invasive manner. It differs from ESS in the way that it fixes the
path length and visitation depth of the scattered photons [37]. The signal obtained has
combined information about morphology and biochemistry of the biological sample.
Its system is based on two spectrometers which are used for illumination and
collection of signals with the help of fibers. A halogen lamp is used as a white light
source. It has been applied in the medical field for breast tissue [38], bronchial tree
[39] and oral mucosa [40].
1.4.4 Near Infrared Spectroscopy
The spectrum in the range 800-2500 nm is known as near infrared (NIR). It penetrates
deep into biological tissues. It probes the chemical bonds of biological molecules
[41]. Differentiation of biological samples on the basis of various functional groups
and chemical bonds is achieved by using the NIR spectrum. It has been applied on
human tissues to grade the various kinds of neoplasia [42].
1.4.5 Fluorescence Spectroscopy
All the tissues contain fluorophores which make the tissue to fluoresce when light is
made incident on it. The fluorescence spectrum contains the information of all the
fluorophores present in the sample which help in differentiation between the two
different groups of samples with varying composition [41], [43]–[45]. In cases where
malignant tissues cannot be identified by white light, fluorescence spectroscopy is
useful. An early diagnosis of laryngeal cancer is a remarkable achievement with this
technique [46].
Introduction
6
1.4.6 Raman Spectroscopy
A non-resonant type of scattering in which the scattered photon has different energy
than that of the incident one is known as Raman scattering. Raman spectroscopy is a
vibrational spectroscopic technique where the interaction of electromagnetic (EM)
waves with a molecule changes its polarizability which results in the Raman bands.
Each band corresponds to a specific energy transition between the vibration energy
levels. A Raman spectrum consisting of these specific Raman bands is known as the
molecular-fingerprint of a target molecule. Raman spectroscopy is a sensitive
technique with high specificity. It is described in details in next chapter. It can be used
effectively for various applications in different fields of science and technology [47]–
[49].
1.5 Thesis Layout
Thesis of the present research work is presented in five chapters. Over view of the
problem addressed in the present research is described briefly in chapter 1 with a
description of DENV infection and its diagnosis. Moreover, brief information
regarding various optical spectroscopic techniques has also been provided along with
Raman spectroscopy and its efficacy regarding diagnosis of DENV infection in the
same chapter. In chapter 2, the first part is dedicated to the description of details about
DENV infection and the sequence of the events that occur during this disease. Then
various most frequently used diagnostic techniques have been described along with
their advantages and drawbacks. In the second part of chapter 2, Raman spectroscopy,
its theoretical and mathematical description on the basis of Classical Physics and
Quantum Physics, related instrumentation and its advantages are discussed. Chapter 3
contains the information regarding the process of samples collection, experimental
setup for Raman spectrum acquisition, statistical methods for model development,
processing of the Raman spectra and biological molecules based analysis of
regression coefficients of the developed model by PLS regression. All the published
articles of the present research work are presented in chapter 4 where experimental
works, analysis, results and discussions are given for NS1 based study [50], IgM
based study [51], IgG based study [52] and lactate based study [53]. Finally, an
overall conclusion of the present research work is given in Chapter 5.
7
2 Diagnosis of DENV Infection and Raman
Spectroscopy
2.1 DENV Infection
DENV infection is a mosquito borne epidemic disease of tropical and subtropical
areas of the world with a high degree of mortality, even in twenty-first century. It is
being investigated with great intensity; however, still there are various unidentified
causing factors and dynamics of this disease. Some of the factors responsible for its
worldwide spread are population growth and transportation of the modern world [54].
Symptoms of DF include flu, headache, diarrhea, loss of appetite, muscle/joint pain,
neurological manifestation, bleeding tendency, thrombocytopenia and skin rash [55].
The DENV is responsible for this disease.
2.1.1 DENV Structure
Encapsulated RNA of DENV, which is 11 kb in length, encodes ten types of proteins.
Three of them are structural proteins while seven are non-structural. One of these non-
structural proteins is known as NS1. It has an important role from diagnostic point of
view. Four sera types of DENV are known as DEN-1, DEN-2, DEN-3 and DEN-4.
They are transmitted by female aedes aegypti mosquito [56]. Unfortunately, there is
no specific vaccine available to control it. It is reported that DENV is not neutralized
by antibodies produced against it by human body immune system. It is an extremely
alarming fact to know that these antibodies may play a negative role by antibody-
dependent enhancement (ADE) of DENV infection [57]–[59].
2.1.2 DENV Strains and ‘Immune Protection’
Initial fever in this disease is categorized as DF while its severe cases are categorized
as DHF and DSS [60]. If a patient is supported with therapy then a study shows that
only 1-5 % cases of DF have resulted in death. Only few of these cases move towards
Diagnosis of DENV Infection and Raman Spectroscopy
8
DHF and DSS where plasma leakage is the first sign. In certain cases this shock leads
to death of the patient [61].
After the bite of a mosquito, it usually takes 4-7 days to show the first sign as
fever. In these initial days the virus replicates itself rapidly and it rises to a certain
level that a biting mosquito can be infected. This stage of high concentration on
DENV in the body is known as viremia [61]. During viremia, NS1 is found abundant
in the serum. The immune system responds to it by producing IgM in 5-6 days from
the day of first mosquito bite. After 7-10 days human body immune system produces
IgG which last for a very long time in the body. If a patient is reported with a high
level of IgG in an acute phase of DF then it is termed as secondary infection [7]–[9].
Presence of IgG in the body ensures protection from future DENV infections. But it
protects only against the serotype of DENV of last infection, so the body still remains
vulnerable to other serotypes. A person with infection of two or more serotypes of
DENV at a time increase the chance of DHF and DSS [62]. There is a hypothesis that
if a body contains antibodies which are cross-reactive and non-neutralizing produced
by an earlier infection, then secondary infection is facilitated by these agents with the
help of Fc (a portion of the immunoglobulin molecule) receptors [63].
2.1.3 Diagnostics Techniques
Virus isolation
It is a usual practice for virus isolation that samples of tissue, blood, serum or plasma
are collected well within 5 days, when viremia occurs. These samples are required to
be cooled at 4-8 oC if the storage period is less than a day; otherwise, i.e. for longer
periods these samples should be cooled well at about -70 oC. For DENV isolation, it
is a usual practice to culture the cells with the help of host cells [64], [65]. Isolation of
DENV was performed by Kimura and Hotta in 1943. Four techniques have been
successfully used in different studies [10]–[13]. According to one of these studies [9]
aedes albopictus mosquito cells were cultured in a minimum essential medium
(MEM). The cells were cultured at 28 °C for 7 days. Culture supernatants were
collected and checked for the presence of dengue virus. The serotypes of the isolated
dengue viruses were determined by RT-PCR [9]. This whole process may take 7-14
days and the results are positive only if the samples are well stored and transported
such that it could not affect the viability of DENV in these samples [65].
Diagnosis of DENV Infection and Raman Spectroscopy
9
This technique has certain issues like high cost, expert staff, sophisticated
equipment, reagent purity and most importantly it takes about 7 days. All these
drawbacks make it unsuitable for a routine-based diagnosis technique in the days of
an outbreak of DENV infection.
Genome detection
RT-PCR method enables the detection of RNA of DENV in serum, plasma, cells,
tissues etc. It takes one day to perform this test. Serotyping can also be done with this
technique [66]. Conventional RT-PCR has been reported to have sensitivity and
specificity 48.4-100 % and 100 % respectively [15], [16], [66]. Sensitivity is low if
the sample is not of acute phase, which is a drawback of this technique. Moreover,
frequent false-positive results, costly equipment and the need of skilled technician
also make it inconvenient to be adopted as a common diagnosis technique for DENV
infection [31], [67].
This technique has been reported to be successfully used in some studies [68],
[69]. A serum sample was used for isolation of RNA. Process of incubation for
reverse transcriptase was done 53 °C for 10 minutes. Then the process of polymerase
chain reaction was conducted in 30-40 amplification cycles.
ELISA based Antigen Detection
As mentioned in section 2.1.1, NS1 appears in seven days of a mosquito bite, so it
becomes the first biomarker of activity regarding replication and viremia of DENV.
Importantly, its concentration is found to be higher if the subject is a potential case of
DHF [31]. Some of the techniques like immuno-histochemical, immuno-fluorescence
assay (IFA) and radio-immuno-assay (RIA) have been investigated for detection of
NS1, but the success rate was very low as compared to ELISA [30].
In the year 2000, Young and his coworkers succeeded in capturing NS1 in acute
phase serum samples with a method that is known as ELISA. It is a very complex
chemical process which has been explained in great detail by K. Bundo K and A.
Igarashi [70]. Firstly, a layer of adsorbed antigen/antibody is formed in the well plate.
This is called coating. Then the blocking is performed and after that the process of
detection is completed. It is important to note here that this process takes place only at
the surface so it is required to wash the well plate after every step so that unbounded
chemical should be removed properly. Dilution is advised to be avoided for better
Diagnosis of DENV Infection and Raman Spectroscopy
10
results. Various layers of antibodies can help in amplification of the final signal so it
is advised to use this technique of multiple layers. Tools for these techniques are
commercially available in kits with ELISA format. They have potentially replaced the
methods of virus isolation and RT-PCR for the screening of DENV infection.
Sensitivity of these kits has been reported to be 94.7-98.3 % and 67.1-77.3 % for
primary infections and secondary infections respectively. However, specificity is 100
% for both type of infections [30]. Low sensitivity of NS1 in case of secondary
infection is a potential drawback as these infections lead to DHF and DSS which is of
prime importance from doctors’ point of view.
ELISA based Antibodies Detection
There are several methods for the detection of antibodies in the serum e.g.
hemagglutination inhibition (HAI) [7], complement fixation (CF) and neutralization
test (NT) [71]. However, these methods are not very commonly used due to several
drawbacks and complexities involved in their procedures. Accurate detection of IgM
and IgG at an earliest possible stage is very useful. Different companies like Standard
Diagnostics, South Korea and Alere, Australia have introduced ELISA based kits for
IgM and IgG detection. These kits are commercially available. Several studies have
been performed on these kits to determine the performance of these kits especially in
secondary infections [16], [72]–[74]. These studies showed that sensitivity and
specificity of these kits for IgM detection ranged in 20.22-99 % and 52-100 %
respectively. Moreover, sensitivity and specificity of these kits for IgG detection
ranged in 78-88.9 % and 63.5-100 % respectively. Cross-reactivity due to Japanese
encephalitis, yellow fever and secondary infections were causing a huge problem in
the reliability of results by these kits as they produce false-positive results quite often
[75]. Frequent use of ELISA based IgM and IgG detection kits around the world is
aimed at discriminating the primary infection from secondary infection.
Virus detection and RT-PCR are not viable solutions for routine based diagnosis
in clinics because these techniques are time consuming, costly and require highly
skilled technicians. ELISA is the technique which is being used as a routine test for
DENV infection due to low cost, quick and robust nature as compared to virus
detection and RT-PCR. However, WHO reports that due to impurities in the reagents
used in ELISA, it results in false-positive and false-negative outcomes [76]. An
ELISA kit is designed to be used for samples in a batch, which means that using this
Diagnosis of DENV Infection and Raman Spectroscopy
11
technique for a lone sample is not feasible economically as the kit is valid to be used
during a very short time after it is opened.
It is a fact that the test that provides highly specific and sensitive diagnostic
results are usually more sophisticated and require costly equipment, highly skilled
technicians and more time. Rapid diagnostic tests are simpler, quicker, robust and
cheap, but their sensitivity and specificity are low. Hence, the sensitivity and
specificity are inversely related with an easy and rapid diagnostic test [77].
2.1.4 Dengue in the Future
It is feared by some of the researchers that the DENV infection has the potential to
increase due to various supporting factors e.g. increased travelling, poverty, poor
hygiene and poor sanitation conditions. Its control, diagnosis, management and
prevention will be the hot topic of future research. It is important to explore all the
factors responsible for DENV infection so that the objectives of WHO about a healthy
world could be achieved [78]–[83].
2.2 Raman Spectroscopy
2.2.1 History of Raman Scattering
Smekal was the first to postulate the scattering of light in an inelastic manner in 1923
[84]. Landsberg and Mandelstam saw unexpected frequency shifts in scattering from
quartz [85] in1928. In the same year, Sir Chandrasekhara Venkata Raman became the
pioneer, who demonstrated it practically by using filtered light of the sun and a
telescope along with his co-worker Krishnan [86]. After a span of two years he was
awarded The Nobel Prize for physics in 1930. This type of inelastic scattering is
named after him as ‘Raman effect’ or ‘Raman scattering’. With the invention of laser
systems in 1960s, this field of research grew rapidly. In 1977 SERS was discovered
and single molecule detection was reported in 1997. Research in this field saw a boom
with the availability of thermo electric cooler (TEC) cooled detectors of high
resolution and grating based monochoromators. Fiber optics helped in the application
of this phenomenon in various fields.
Diagnosis of DENV Infection and Raman Spectroscopy
12
2.2.2 Raman Shift
A Raman spectrum is obtained by recording the intensity of scattered photons against
their Raman shift in cm-1. Raman shift for each photon can be calculated by Eq. 2-1
where 𝜆𝑖 and 𝜆𝑠 are the wavelengths of incident and scattered photons respectively in
centimeters.
𝑅𝑎𝑚𝑎𝑛 𝑠ℎ𝑖𝑓𝑡 (𝑐𝑚−1) = 1
𝜆𝑖−
1
𝜆𝑠 2-1
A representative Raman spectrum is shown in Fig. 2-1.
Figure 2-1 A typical Raman spectrum.
2.2.3 Classical Description
Brief introduction of Raman scattering is given in section 1.4.6. Classically, a di-
atomic molecule can be assumed to be made of two masses m1 and m1 attached with
each other by means of a spring with spring constant K as shown in Fig. 2-2. Here,
masses m1 and m2 represents two atoms while spring represents the chemical bonding
between these two atoms. Displacement of masses m1 and m2 are represented by x1
and x2 respectively.
Diagnosis of DENV Infection and Raman Spectroscopy
13
Figure 2-2 Diatomic molecule as a mass on a spring.
Hooke’s law can be written for this type of system as
𝑚1𝑚2
𝑚1 + 𝑚2(
𝑑2𝑥1
𝑑𝑡2+
𝑑2𝑥2
𝑑𝑡2) = −𝐾(𝑥1 + 𝑥2) 2-2
Here (m1m2/[m1+m2]) represents the reduced mass of the system which can be
represented by μ. Moreover, (x1+x2) represent the total displacement and it can be
replaced by q. Now the above equation can be rewritten in simple form as
𝜇 (𝑑2𝑞
𝑑𝑡2) = −𝐾𝑞 2-3
This second order differential equation can be solved for q, which yields a
solution as
𝑞 = 𝑞𝑜𝑐𝑜𝑠(2𝜋𝜈𝑚. 𝑡) 2-4
In this solution, νm represents the vibrational frequency of the molecule shown
in Fig. 2-2. Its value can be determined by
𝜈𝑚 =1
2𝜋√
𝐾
𝜇 2-5
Equation 2-3 shows that the vibration of a molecule follows a cosine pattern
whereas equation 2-4 depicts that the strength of a chemical bond among the atoms is
directly proportional the frequency of vibration. Reduced mass is however inversely
related to the vibrational frequency. It can be concluded that molecules of different
chemicals will have their own characteristic pattern of these vibrations. Polarizability
of a molecule α is another important term. This property of a molecule shows the
tendency of a molecule to polarize its electronic cloud under the influence of external
electric field. Polarizability is a function of displacement q. In case of Raman
scattering the external electric field is provided by the electric field of an excitation
source of light which is usually a laser beam. The dipole moment P induced due to an
applied field E in a molecule with polarizability α is given by
𝑃 = 𝛼𝐸 2-6
As the electric field is provided by the incident EM wave, in which E oscillates
in a cosine pattern with amplitude Eo and frequency νo, given by
Diagnosis of DENV Infection and Raman Spectroscopy
14
𝐸 = 𝐸𝑜cos(2𝜋𝜈𝑜. 𝑡) 2-7
This equation is used in equation 2-6 to get
𝑃 = 𝛼 𝐸𝑜𝑐𝑜𝑠(2𝜋𝜈𝑜. 𝑡) 2-8
Using an approximation which is known as small amplitude approximation, it is
possible to write α as a linear function of q
𝛼 = 𝛼𝑜 + (𝜕𝛼
𝜕𝑞)
𝑞=0
. 𝑞 + ⋯ 2-9
Making use of this expression for α in equation 2-8
𝑃 = 𝛼𝑜 𝐸𝑜𝑐𝑜𝑠(2𝜋𝜈𝑜. 𝑡) + (𝜕𝛼
𝜕𝑞)
𝑞=0
. 𝑞𝑜𝑐𝑜𝑠(2𝜋𝜈𝑚. 𝑡) 𝐸𝑜𝑐𝑜𝑠(2𝜋𝜈𝑜. 𝑡) 2-10
First part on the left side shows the elastic scattering known as Rayleigh scattering,
while the second term represents inelastic scattering which is known as Raman
scattering. Interestingly the last term has two terms of cosine functions which are
being multiplied. This product can be converted into a sum of two cosine terms by
using a trigonometric formula 2.Cos(A).Cos(B)=Cos[(A+B)/2]+Cos[(A-B)/2], which
yields
𝑃 = 𝛼𝑜 𝐸𝑜𝑐𝑜𝑠(2𝜋𝜈𝑜. 𝑡) +1
2(
𝜕𝛼
𝜕𝑞)
𝑞=0
𝑞𝑜𝐸𝑜 . 𝑐𝑜𝑠(2𝜋{𝜈𝑜 − 𝜈𝑚}. 𝑡)
+1
2(
𝜕𝛼
𝜕𝑞)
𝑞=0
𝑞𝑜𝐸𝑜. 𝑐𝑜𝑠(2𝜋{𝜈𝑜 + 𝜈𝑚}. 𝑡) 2-11
Here second term on the left side of the equation represents the oscillating electric
field with frequency which is νo-νm, i.e. less than the incident photon. This represents
Stokes Raman scattering. While, the third term on left hand side represents the
oscillating electric field with frequency νo+νm, i.e. greater than that of the incident
photon. This represents the anti-Stokes Raman scattering.
It is important to note here that the condition which is found to be necessary for
Raman scattering is ∂α/∂q≠0. This condition says that the change of position of atoms,
in a specific vibrational mode, must result in the change of polarizability. In order to
understand this point, consider a diatomic molecule; say AB, in Fig. 2-3. Let L be the
length of bond A-B in equilibrium position. Assume that the maximum displacement
of its atoms from the mean position in a specific vibrational mode is qo.
Diagnosis of DENV Infection and Raman Spectroscopy
15
Figure 2-3 Bond length of a diatomic molecule during a vibration.
Figure 2-4 Polarizability as a function of vibrational displacement about
equilibrium.
At the position of maximum compression the atoms of the molecule are so much
close to each other that they are least affected by the externally applied electric field
i.e. incident EM wave. As these atoms start to go away from each other the value of
polarizability α starts to increase and reaches to its equilibrium value αo when atoms
reach back to their equilibrium position. It moves on to maximum elongation and at
that point the molecule can easily be perturbed by externally applied electric field.
Here α has maximum value. In Fig. 2-4 the values of α against the separation of atoms
Diagnosis of DENV Infection and Raman Spectroscopy
16
is shown for understanding of the condition for Raman scattering to occur. The
Raman scattered light will have frequencies νo-νm and νo+νm. This condition can be
considered to be the selection rule of Raman scattering α.
2.2.4 Quantum Description
According to quantum mechanics a molecule has quantized electronic, vibrational and
rotational energy levels. These energy levels are shown in Fig. 2-5.
Figure 2-5 Energy level diagram.
Atoms of a diatomic molecule vibrate about the mean position and they behave
just like a harmonic oscillator which has quantized energy levels i.e. Ej=hνj[j+(1/2)].
Here j represents the quantized vibrational energy level as shown in Fig. 2-6.
Therefore, it can be said that the difference of energies of levels E1 and E0 is ΔEm. i.e.
νm= ΔEm /h [47]–[49]. According to Boltzmann distribution function in a collection of
molecules than will be a certain number density of molecules in a state with j=0
which will be more than the number density of the state with j=1 and so on. This is
linked with the temperature of the molecules. Due to interaction of a photon with a
molecule, a transition in the vibrational energy level takes place and Raman scattering
Diagnosis of DENV Infection and Raman Spectroscopy
17
takes place as a result of this transition. The EM wave of this photon sets up an
oscillation g dipole moment and molecule-photon complex jumps to a virtual energy
level. Energy level of this virtual state is not equal to any electronic state, but is
greater than the vibrational level, so the molecule remains in the ground electronic
state. During this interaction some quanta of energy is kept by the molecule which
corresponds to the vibrational mode of that molecule and surplus energy is taken
away by the Raman scattered photon as shown in Fig. 2-6.
Figure 2-6 Conservation of energy for Raman scattering (Stokes).
If a molecule jumps from lower vibrational level to a higher vibrational level, then the
energy of Raman scattered photon will be h(νo-νm) i.e. Stokes Raman scattering (Δj=
1). Whereas, if a molecule jumps from higher vibrational level to lower vibrational
level, then the Raman scattered photon will the carry the energy h(νo+νm) i.e. anti-
Stokes Raman scattering (Δj= -1). As depicted by the factor in Boltzmann distribution
of molecules, it is evident that the number density of molecules in lower vibrational
energy level will be more than the number density in higher vibrational energy level,
it can be said that the Stokes Raman scattering and anti-Stokes Raman scattering will
take place simultaneously but the intensity of latter will be lower comparatively [51]–
[53]. The same phenomenon is shown in Fig. 2-7, along with Raleigh scattering, IR
absorption and fluorescence to have a comparative understanding with Raman
scattering.
Diagnosis of DENV Infection and Raman Spectroscopy
18
Figure 2-7 Energy level diagram for Raman scattering.
2.2.5 Raman Signal Intensity
Typically, out of 107 incident photons only 1 photon is scattered inelastically. This
gives an idea that how much low intensity of Raman signal can be detected in
collected scattered light. It is important to have a view of the parameters that
determine the strength of Raman signal. The relation of intensity of Raman signal ФR
with different parameters is given by
𝛷𝑅 ∝ 𝜎(𝜈𝑒𝑥)𝜈𝑒𝑥4 𝐸𝑜𝑛𝑖𝑒
−𝐸𝑖𝑘𝑇
Here νex is excitation frequency, σ(νex) is Raman scattering cross-section for a
particular wavelength, Eo, is the irradiance of the incident beam, ni is the number
density of the target sample and the last exponential term is Boltzmann factor for state
i. Raman scattering cross-section is compared with cross-sections of different
processes in the Table 2-1.
It is important to note that the intensity of Raman signal is inversely
proportional to the fourth power of wavelength of the incident beam as shown in
Table 2-2. By using a shorter wavelength and higher power more intense Raman
signal can be recorded from the same concentration of scatterers in a sample.
Diagnosis of DENV Infection and Raman Spectroscopy
19
Table 2-1 Cross-sections of most common optical processes [87].
Cross-section of process σ(νex) (cm2)
Surface enhanced resonance Raman scattering 10-15
Surface enhanced Raman scattering 10-16
Absorption of UV 10-18
Emission of fluorescence 10-19
Absorption of IR 10-21
Resonant Raman scattering 10-24
Rayleigh scattering 10-26
Raman scattering 10-29
Table 2-2 σ(νex) of Raman scattering for CHCl3 at different incident wavelengths
[87].
λex (nm) σ(νex) ( x 10-28 cm2)
532.0 0.66
435.7 1.66
368.9 3.76
355.0 4.36
319.9 7.56
282.4 13.06
2.2.6 Quantitative and Qualitative Analysis
From 1928 to the end of the last century, the components used for Raman
spectroscopy were neither cheap not that much efficient as they are at the moment.
This led to a delay in the application of Raman spectroscopy for the most part of the
last century in various fields. Now with stable monochromatic laser systems in the
form of diode lasers, high resolution TEC cooled CCD detectors, fiber optics based
probes and more intelligent and programmable applied multivariate statistical analysis
methods; not only qualitative analysis but also quantitative analysis can be performed.
Hence, by using Raman spectroscopic analysis, the concentration of particular specie
in a sample can be determined.
Diagnosis of DENV Infection and Raman Spectroscopy
20
2.2.7 Recent Advances in Raman Spectroscopy
Surface Enhanced Raman Spectroscopy (SERS)
SERS is a technique in which the strength of the Raman signal is enhanced to the
order of 1014-1015. The use of specially constructed nano-surfaces and nano-particle is
getting famous among the researchers who are interested in detection and analysis of
chemical present in trace amounts in samples. Gold and silver are being used for this
technique while copper and aluminum is also under investigation. Kits for SERS are
commercially available as well. It is interesting that the exact enhancement
phenomenon of this technique is still a topic of debate. However, chemical and
electric filed enhancements are two well-known factors considered responsible.
Electric field theory takes into account the mechanism of surface plasmon resonance
whereas; chemical theory explains it on the bases of charge-transfer-complexes [88].
Coherent Anti-Stokes Raman Spectroscopy (CARS)
It uses the interaction of a pump laser beam with frequency ωp and a Stokes laser
beam with frequency ωs. Frequency of CARS signal is given by 2ωp- ωs. When the
beat frequency of a pump beam and Stokes matches with the frequency of a particular
Raman band, then it results in an enhanced anti-Stokes Raman signal, which is known
as CARS [89]. It has some advantages which can exploit the signal from a particular
Raman band of a molecule of interest from the target sample for diagnosis or analysis.
Setup for this type of spectroscopy is not simple to be established.
Resonance Raman Spectroscopy (RRS)
In this type of Raman spectroscopy, a tunable dye laser is used to specifically target
the transition of interest out of all the available transitions in a large biological
molecule where fluorescence competes and subdues the strength of the Raman signal
of interest. Due to intentional involvement of resonance between an incident laser
frequency and the frequency of target transition it is known as resonant Raman
spectroscopy [90].
Polarized Raman Spectroscopy (PRS)
It utilizes a specialized setup where an excitation beam with a specific polarization is
made incident on the sample. The Raman scattered light is passed through filters to
Diagnosis of DENV Infection and Raman Spectroscopy
21
collect the Raman signal of a specific polarization at the detector. This technique can
provide us information about orientation of the bonds in a molecule [91].
2.2.8 Diagnostic Applications
In recent years Raman spectroscopy has been used in medical field, especially for the
diagnosis of various diseases, including infectious diseases and cancer. Importantly,
few of them were in-vivo studies [92]–[94]. At University Hospital, Groningen, a
group of researchers have shown that Raman spectroscopy can clearly differentiate
the three type of cell layers [95]. Excised breast tissues were successfully classified by
Raman spectroscopy according to the stage of cancer progression which correlated
very closely to those of clinical results obtained by histo-pathological methods [96]–
[98]. In a recent research by a group at National Institute of Lasers and
Optronics (NILOP), it has been demonstrated that Raman spectroscopy can be
used to screen female BRC cases from whole blood samples [99]. A similar
kind of research was carried by another group which also showed promising
results for further research in this field [100]. Barrett's esophagus was investigated
for changes that are accompanied by the carcinogenesis by Raman spectroscopy
[101]. Thyroid cell lines were also successfully screened for malignancy by using
Raman spectroscopy [102]. Hepatitis C is infectious disease which is screened by
Raman spectroscopy by a group at NILOP recently [24]. Moreover, Raman
spectroscopy is also used by that group for screening of nasopharyngeal cancer in the
human blood sera [103]. Raman spectroscopy can become a suitable detection probe
with lab-on-chip (LOC) devices to produce an efficient clinical diagnostic test [104],
[105].
A study reported the detection of NS1 by using SERS. It is reported to be
highly sensitive technique which is capable of distinguishing between NS1 of Zika
virus and DENV [106]. Dr. Fengwei Bai’s group has used gold-nanoparticles to detect
antibodies against DENV [109]. They have reported it to be a quick and sensitive
assay as compared to RT-PCR [107]. A group of scientists at NILOP, Pakistan is
working on diagnosis of DENV infection by Raman spectroscopy at wavelengths of
785 nm and 532 nm. Various multivariate statistical classification techniques e.g.
principal component analysis (PCA), PLS, support vector machine (SVM) and
Diagnosis of DENV Infection and Raman Spectroscopy
22
random forest (RF) analysis are applied on Raman spectra, and have reported good
results.
2.2.9 Advantages
Some of the advantages of Raman spectroscopy are described here briefly.
No sample preparation: A Raman spectrum can be acquired from a biological
sample without any specific physical or chemical preparation. This allows us to use
this technique in-vivo.
Non-destructive: With proper wavelength selection and power optimization, this
technique can be used for non-destructive application, which is prime objective for
analysis of any biological sample, from the medical point of view.
Aqueous samples: IR spectroscopy is influenced intensely due to the presence of
water which is an essential part of biological samples, but in case of Raman
spectroscopy, it does not affect the Raman signal and is a better choice for analysis.
Specificity: The Raman spectrum of a chemical will have specific bands for that
particular chemical only. Hence a specific chemical has a specific Raman spectrum.
This spectrum is termed as a molecular fingerprint of that particular compound.
Organic and inorganic samples: The Raman spectroscopy can analysis both types of
samples no matter whether they are organic or inorganic.
Wide Concentration Range: With recent advancements it has been made possible to
detect the presence of a single molecule in the sample. Moreover, there is no need to
dilute the sample if a sample has a very high concentration of certain chemical.
Windows compatibility: Raman spectroscopy can be used to analyze the samples,
which are contained in common transparent containers.
Quick: It does not require any sample preparation and any sort of time wasting
process which make it a quick method to obtain a Raman spectrum and quickly
analyze it with modern day laptops with the help of available software. Real time
analysis is a huge plus for its application during a chemical process and surgery of
infected areas especially tumor removal.
Diagnosis of DENV Infection and Raman Spectroscopy
23
Mixture of Molecules: If a mixture of large molecules is excited by a laser beam then
then the recorded Raman spectrum will contain Raman peaks of all the molecules
present in that sample. Therefore, information about all the molecules can be analyzed
by a single Raman spectrum. It makes the study of complete picture of the biological
samples easier to characterize.
2.2.10 Disadvantages
Low excitation probability: As described in previous sections that Raman scattering
cross-section is of the order 10-29 which is extremely low as compared to other
processes, so it requires very high grade optics and TEC cooled detectors to collect as
much Raman signal as possible.
Fluorescence: During the process of detection of a Raman signal, the fluorescence
signal is also recorded. As the strength of the fluorescence signal is more than Raman
signal, therefore it causes problems for purely Raman scattering based spectrum
acquisition and analysis. At higher wavelengths, the fluorescence signal decreases,
but it causes a decrease in Raman signal as well because of λ-4 dependence.
Raman signal at longer distances: The Raman signal is extremely hard to collect at
longer distance as compared to the fluorescence signal. So this technique can only be
applied efficiently for closely focused locations.
Overlapping of Peaks: When a sample contains large number of molecules, then it is
highly likely that some of the Raman peaks which lie very close to each other in the
spectrum may overlap with each other. Such peaks are either suppressed or appear as
shoulder peaks, which makes the analysis somehow difficult.
2.3 Summary
A lot of efforts have been made for the diagnosis of DENV infection. Chemicals
methods have been developed improved and are being used for screening purpose all
over the world at the moment. However, sensitivity and false-positive results are still
a big concern and it invites the researchers to develop more sensitive and accurate
diagnostic techniques, especially at an earliest possible stage. These new methods
should be cheap because this disease is affecting developing countries very badly. The
time factor is also important for two reasons. Firstly, quick screening of a suspected
Diagnosis of DENV Infection and Raman Spectroscopy
24
person is important because during the outbreak of DENV infection, the inflow of
suspected subjects at health care units increases more than its capacity due to waiting
for the results of their tests. Secondly, DENV infection is not cured, but is managed to
sustain the good heath of infected person with intake of various medicines and a
prescribed diet plan [108]. The earlier a person is diagnosed with DENV infection, the
more are the chances for his betterment and survival. New diagnostic techniques
should also be simpler to be used so that it should not require a highly skilled
technician for its operation.
All these needs can potentially be met if Raman spectroscopy could be applied
for this cause. Due to several advantages, Raman spectroscopy is a potential candidate
to be developed as a technique for the diagnosis of DENV infection. With extremely
low running cost, no chemical consumption, no sample preparation, software based
automatic acquisition of Raman spectrum and model based automatic diagnosis of
DENV infection in a quick way will help the affected developed countries a lot in
controlling the mortality rate.
25
3 Materials and Methods
3.1 Collection of Samples
There are different body fluids which may have been used for this screening
study but the choice of serum was made to avoid the problem of thermal degradation
of sample due to laser excitation. Serum has almost all the biological molecules which
may have any role in DENV infection. Moreover the sample of serum is much easier
to acquire, handle and store for analysis as compared to saliva, urine, tears etc. In
short, the serum is rich in biological molecules, easily acquired and handled. Samples
of blood serum were collected from Rawalpindi Medical College (RMC), Rawalpindi,
Pakistan and its allied hospitals. A number of studies have been conducted during the
present research work and for each case, a different set of samples were collected and
analyzed. A number of hospitals played their role in the present research e.g. Nuclear
Medicine Oncology & Radiotherapy Institute (NORI) Islamabad, Meyo Hospital
Lahore and PAEC Hospital Islamabad. Blood samples were collected from suspected
subjects, who were reported with symptoms of DENV infection, before any clinical
confirmation and medication. A sample of 3 ml non-heparinized blood from arm-pit
was drawn into a tube, recommended for serum extraction. Serum was extracted by
means of a centrifuge machine at RMC. Serum was poured into two centrifuge tubes,
one for ELISA based analysis of NS1, IgM and IgG while other for Raman
spectroscopy at NILOP. From sample extraction phase to the final stage of sample
disposal, it was ensured that all the safety guidelines of the National Institute of
Health (NIH) [109] must be followed in letters and spirit.
3.2 Experimental Setup
A Raman system is composed of three basic units, which are: an excitation source of
light, a spectrometer and a detector. A light source is usually a laser of appropriate
wavelength and adjustable power. In addition to these basic units, microscope and
fiber optics are also used. Fiber optics is used for guiding excitation light to the
sample and collection of Raman signals from the sample.
Materials and Methods
26
Figure 3-1 Schematic diagram of a typical fiber optic probe based Raman
system.
A filter is used to stop Rayleigh signal usually. With the help of grating
spectrometer it is made possible that the photons of different wavelength fall at
different pixels of a TEC cooled CCD detector according to calibration. A laptop is
interfaced with CCD to collect, record and plot the Raman spectrum. A schematic
diagram of a typical Raman system is shown in Fig. 3-1.
The Raman system used in the present research work is shown in Fig. 3-2. It
was manufactured by Agiltron, USA with commercial name PeakSeeker Pro-785. It
consists of a diode laser @785 nm wavelength, an optical fiber probe, a high pass
filter grating based monochromator and TEC cooled CCD detector. An optical fiber
probe contained an excitation fiber as well as several collection fibers. A high-pass
filter was used to collect the Stokes Raman scattering signal only. A TEC cooled
CCD detector was used with controlled temperature at -20 oC to minimize the thermal
current in the detector. The system was coupled with a computer interface. The
graphical user interface (GUI) of software was installed at the computer which was
used to operate the Raman system. Laser power can be set in the range of 5-300 mW
and integration time in the range of 1-30 seconds. Focusing of excitation and
collection was achieved by coupling the probe into an integrated microscope (RMS-
785 by Agiltron, USA).
Materials and Methods
27
Figure 3-2 Experimental setup of the Raman system.
A camera attached on top of this microscope provided the option of visualizing
the focusing point as well as capturing the image of focused surface before and after
the recording of a Raman spectrum to access any sort of change or thermal
degradation of a sample for power optimization. A 10x objective was used throughout
the present research work. A metallic substrate sheet of aluminum was preferred over
other substrates like glass and quartz due to lower background and fluorescence
contribution. Raman spectra were recorded in the range of 300-1800 cm-1 with a
resolution of 10 cm-1. The Raman spectrum was calibrated by using a silicon wafer
with characteristic peak at 520 cm-1.
3.3 Preprocessing Methods
All the molecules present in the sample produce their characteristic Raman
fingerprint. This spectral overlapping of nearby bands makes it difficult to visually
analyze them qualitatively and quantitatively. In addition, the presence of natural
fluorophores in the sample produces fluorescence background that makes the
characterization of the biological samples more complex and difficult. A multivariate
model for the discrimination analysis can help us minimize such difficulties [110].
PLS regression is a multivariate method which enables us to discriminate even minute
level of variations found in the spectra of biological samples. This technique is
Materials and Methods
28
preferably applied on preprocessed spectral data. Preprocessing is used to eliminate
fluorescence background, electronic noise and the substrate effects to an optimum
level. Preprocessing methods are employed by MATLAB (Mathworks 2009a), prior
to their use in the development of a multivariate model. User friendly GUI is
developed in the MATLAB (Mathworks 2009a) environment for carrying out all
preprocessing, regression model development and analysis.
Denoising: A typical Raman spectrum is recorded in raw form in Fig. 3-3(a). First of
all, electronic noise from the Raman spectra has been removed. The denoising of
spectra is performed by using ‘wden’ function in MATLAB. Wavelet decomposition
and reconstruction method [111] is employed by this function to eliminate noise. In
this method, a spectrum is decomposed into orthogonal wavelets of level 10. A soft
thresholding is used based on the Stein's principle of unbiased risk. The spectrum is
reconstructed by using wavelet coefficients. Stein's unbiased risk estimator
reduces the background in the spectrum; as a result, pixel intensity levels below the
threshold are minimized.
Smoothing: In the second step, denoised spectra were smoothed using a digital
moving average filter, Savitzky-Golay [112], [113]. It was applied over a span of
seven points with 4rd order polynomial fitting. The combination of these smoothening
functions removes the noise more efficiently by preserving the Raman bands in the
spectra. A denoised and smoothed Raman spectrum is shown in Fig. 3-3(b)
Baseline correction: In the third step, fluorescence background is removed by
baseline correction using ‘msbackadj’ function with a window of 200 cm-1 and a
polynomial of 4th degree for estimation of the baseline [114], [115]. It ensured that
Raman features narrower than 200 cm-1 were preserved, while wider ones are
considered as the fluorescence background and therefore filtered out. A baseline
corrected spectrum is shown in Fig. 3-3(c).
Normalization: Finally, the resulted Raman spectrum is vector normalized as shown
in Fig. 3-3(d). For development of multivariate model, a suitable spectral range is
also selected. All these processes were employed automatically by a code written in
MATLAB (Mathworks 2009a) to ensure the true, unbiased and robust nature of
preprocessing. Finally, the processed spectra were employed as trainee data set in the
PLS regression algorithm for the development of model.
Materials and Methods
29
Figure 3-3 Preprocessing of the Raman spectrum from a raw spectrum to final a
preprocessed Raman spectrum is shown here: a: Raw Raman spectrum, b:
denoised and smoothed Raman spectrum, c: baseline corrected Raman
spectrum, d: vector-normalized Raman spectrum.
3.4 Statistical Analysis
3.4.1 Multivariate Analysis Techniques
Statistical techniques are based on the method of machine based learning from a set of
data (variables) with known results (responses). This data is known as trainee data.
Regression is applied on this data to develop a vector of regression coefficients which
is known as beta vector or regression vector of the model. There are several
techniques which are in use for analysis and prediction of responses by using
variables of unknown samples. A number of multivariate statistical methods are used
e.g. PCA, PLS regression, artificial neural networks (ANN), logistic regression, SVM,
random forest analysis, random forest regression, etc. During the present research
work PLS is mainly implemented, however, SVM and random forest regression have
also been used to assess their usefulness.
3.4.2 PLS Regression
PLS regression predicts responses (Y) from variables (X) [116]–[118]. In the present
study, variables are the intensities of Raman bands of a spectrum of all the samples
Materials and Methods
30
placed in a matrix X such that each row represents the Raman spectrum of a sample.
Thus a matrix X with order n x p will contain data of n samples with p Raman bands.
Here p depends on the number of pixels of CCD used for recording of a spectrum,
which are 1024 in number for PeakSeeker Pro-785.
Figure 3-4 Mathematical description of PLS regression.
The responses Y are the elements of a column matrix where the clinical result
of each sample in matrix X is correspondingly placed in a matrix Y. It can be either
the value of AI of IgM/IgG or a value ‘0’/’1’ to show clinically positive or negative
result of a sample. In order to correlate the responses (Y) with variables (X) a
regression vector is needed as shown in Fig. 3-4. The principal components (PCs) are
determined which contribute to the majority of the variance found in the data. Only
few of these PCs are used to develop the regression vector such that they are capable
enough to predict the responses with the help of variables for every sample of the
trainee data set [119]. This objective is achieved through a process of decomposition
of X and Y according to equations 3-1 and 3-2.
X
t
XX ELSX 3-1
Y
t
YY ELSY 3-2
SX and SY represent the projections of X and Y respectively. Whereas LX and LY
represent the loadings of X and Y. Regression vector correlates X and Y in such a
Materials and Methods
31
way, that a minimum number of PCs are utilized to predict the responses of trainee
data set as accurately as reasonable achievable with over fitting.
Model development
A GUI is used to develop the PLS model. All the spectral data along with their
clinical results is loaded in the GUI. The number of samples to be used in testing data
set is provided to this GUI as an input parameter. It automatically selects the specified
number of samples, in a random manner, to be used for testing of the developed
model. Spectral data of remaining samples, known as trainee data set, is used for
development of PLS model. This segregation of the samples into trainee and testing
data set in a random, unbiased and automatic manner ensures unbiased nature of the
developed model. Numbers of PCs to be used are chosen by using the methods of
Kaiser, Scree and Parallel factor. Firstly, it is aimed at ensuring that the number of
PCs should be less than one third of the number of samples of positive/control group
(with the lowest number of samples). Secondly, it is done to avoid over-training of the
model. Two plots are used for these purposes which are drawn with the help of eigen
values of the data and root mean square of error [120].
Model Evaluation
Only that model is accepted for further testing phase which gives the best
performance according to leave one out (LOO) cross validation method. In this
method data of one trainee sample is left out of the model development and the model
is developed to predict the left out sample. This prediction is plotted on y-axis in the
calibration plot where its clinical result is taken along x-axis. This sample is placed
back into the trainee data set and the process is repeated for all the samples one by
one. In this way a curve is obtained, which is known as the calibration curve. From
these values of predicted responses by model and the corresponding clinical
responses, a correlation coefficient known as R-square (r2) value is calculated. An
ideal value of r2 is 1, whereas its value around 0.9 is considered acceptable for a good
prediction model [121]. Root mean square error in cross validation (RMSECV) is
also calculated. Standard deviation (SD) in error is another parameter which is
calculated to determine the goodness of fit for the model [122].
Materials and Methods
32
Model Testing
After the successful development of a PLS regression model, it is passed through a
testing phase to evaluate its performance for those hidden samples of testing data set,
which were not included in the phase of developing and optimizing of the model. The
Raman spectrum of each of the sample in testing data set is multiplied in a specific
manner with the regression vector and it yields a value of prediction. A predicted
value above cut-off is taken as predicted positive, whereas a predicted value below
cut-off is declared predicted negative. Now, on the basis of these predictions false-
positive, true-positive, false-negative and true-negative results were calculated. These
results are used to determine sensitivity, specificity and accuracy of the developed
model.
Figure 3-5 Area under ROC curve and its interpretation.
Sensitivity is calculated by dividing the number of correct positive prediction by
the total number of positive samples in the testing data set. Specificity is calculated by
dividing the number of correct negative predictions by the total number of negative
samples in the trainee dataset. Accuracy is determined by dividing the total number of
Materials and Methods
33
correct predictions by the total number of predictions made. Root mean square in
error of predictions (RMSEP) is also calculated. Similarly, the standard deviation of
error in predictions is also calculated to assess the performance of the model
according to the criteria that it should as low as possible around zero [121], [122].
A curve is drawn by calculating the true-positive rate against false-positive rate
at different cut-off values starting from lowest to the highest value according to
clinical responses. This curve is known as receiver operating characteristic (ROC)
curve. A typical ROC curve along with performance determining criteria is shown in
Fig. 3-5. Ideally, the area under ROC curve (AUC) should be 1 but a value above 0.8
is reasonably good for a prediction model [123]–[125].
3.5 Molecular Analysis
After successful development and testing of a reasonably good acceptable model, it is
important to translate these statistics based results into disease related bio-molecular-
based results. To achieve this goal, it is important to know the meaning and
importance of highly positive regression coefficient and highly negative regression
coefficients. A regression coefficient against a corresponding Raman shift with a
considerably high value on percent base means that the concentration of the biological
molecule, associated with that particular Raman shift, increases as the severity of the
disease increases. Similarly, a regression coefficient against a corresponding Raman
shift with considerably negative value on percent base means that the concentration of
the biological molecule, associated with that particular Raman shift, decreases as the
severity of the disease decreases. Assignment of Raman bands to particular biological
molecules are done with the help of a database available in literature [126], [127]. It is
then tried to support these finding based on the regression model with other clinical
studies performed to determine the role of such molecules in relation to disease under
consideration. Molecular analysis has been summarized in chapter 4.
34
4 Results and Discussions
To investigate the possible use of Raman spectroscopy as a diagnostic tool for
DENV infection in human blood sera, it was decided to start the Raman spectroscopy
based screening of NS1 positive and negative samples. Results and discussion is
given in section 4.1. Then it was intended to use Raman spectroscopy for screening of
sera samples on the basis of antibody index (AI) of IgM and IgG quantitatively.
Results and discussions of these two studies are presented in sections 4.2 and 4.3
respectively. Moreover, laser @532nm was used to investigate the possibility of
lactate as a potential biomarker for diagnosis of DENV infection, results and
discussions of this part is presented in section 4.4.
4.1 NS1 based Screening
As discussed chapter 1 and chapter 2, in RNA of DENV produces NS1. In the present
study, Raman spectra of NS1 positive and NS1 negative sera were used with a PLS
regression routine to develop a multivariate model for the optical screening of NS1
positive samples in the samples which were suspected of DENV infection. In total
218 blood sera samples from subjects of different ages and genders have been used.
Among all these samples, 95 were NS1 positive and 123 were NS1 negative. For
model development 178 samples were used which contained 80 NS1 positive and 98
NS1 negative samples. For testing of the developed model 40 samples were used.
According to clinical test based on ELISA, 15 samples were NS1 positive and 25
samples were NS1 negative. All the process of sample collection, Raman spectrum
acquisition and preprocessing was performed as described in sections 3.1 - 3.3.
4.1.1 Results and Discussion:
Eight number of PCs were used for the development of PLS regression model. Raman
spectra of all the samples of NS1 positive and NS1 negative group are recorded in
Fig. 4-1. It is a patched area type graph where green color shows NS1 negative group
while red color shows NS1 positive group. Average plot of each group is also shown.
Results and Discussions
35
The model was duly calibrated and the calibration curve produced is shown in Fig. 4-
2. Value of r2 was determined to be 0.9, which is quite promising. Values of
RMSECV and SD for cross-validation were found to be 0.15 each, authenticating the
predictions made by the model. The results of unknown suspected samples were also
plotted against their clinical results in Fig. 4-2. Importantly, the region between o.4 to
0.6 was declared as grey region where the result of a predicted sample is termed as
inconclusive. Value of RMSEP and SD for testing data set was determined to be 0.2
each. Accuracy, sensitivity, specificity and AUC were determined to be 100 %, 100
%, 100 % and 1 respectively.
Figure 4-1 Raman spectra of sera samples used in NS1 based screening study.
Results and Discussions
36
Figure 4-2 Calibration curve of model for NS1 based screening.
Figure 4-3 Regression coefficients of PLS model for NS based screening.
Results and Discussions
37
Important regression coefficients at corresponding Raman bands are shown in Fig. 4-3
where values were strongly positive or strongly negative. A list of important
regression coefficients along with the role of their associated biological molecules
have been given in Table 4-1 and Table 4-2. Few of these molecules have been
discussed here for their role with respect to DENV infection.
Positive regression coefficient at 776 cm-1 depicts that phosphatidylinositol
[128] is found in higher concentration in NS1 positive subjects as compared to
negative. Its high concentration is supported by the fact that expression of NS1
takes place at the surface of infected cells [129] and human defense system
targets NS1 in response to infection. Cellular glycosyl-phosphatidylinositol is
used by DENV for signal transduction capacity as a result of binding of NS1
to specific antibody [130].
Positive regression coefficient at 1127 cm-1 is reported for Raman band of
ceramide [131] whose level is reported to be high [132] in NS1 positive
samples.
At 736 cm-1 a negative correlation exists which is reported to be one of the
three characteristic Raman band of thiocyanate [133]. It is a potentially useful
therapeutic agent with host defense and antioxidant properties [134].
At 1454 cm-1, the positive correlation exist for the Raman band of protein
which has a structural role [135] and the present study shows its higher
concentration in NS1 positive group where structures of viruses are
synthesized.
At 1045 cm-1, proline [136] is reported to have high concentration in case of
DENV infection [132].
Raman bands at 1224, 1254, 1273 and 1283 cm-1 are characteristic Raman
bands of amide-III [127] which showed positive correlation with NS1 positive
group.
Results and Discussions
38
Table 4-1 Positive regression coefficient obtained by PLS model for NS1 based
screening along with molecular description.
Raman
shift
Strength of
regression
coefficients
(%)
Molecular assignments
1174 86.4 DNA bases and protein
789 70.2 C5-O-P-O-C3 phosphodiester bonds in DNA
1283 62.4 Differences in collagen content, amide III
776 46.4 Phosphatidylinositol
1224 42.3 PO2− in nucleic acids and amide-III (β sheet structure of
protein)
1503 42.1 NH3
1454 38.0 Protein and phospholipids
1127 37.7 Protein and ceramide
1273 35.4 DNA/RNA bases and amide-III (proteins)
1254 33.8 DNA/RNA bases and amide-III (proteins)
1045 37.7 Proline
1363 33.1 Tryptophan
608 29.9 Cholesterol
942 29.3 Skeletal modes (polysaccharides, amylose, amylopectin)
1344 28.6 Protein
1466 27.6 CH deformation (DNA/RNA and proteins and lipids and
carbohydrates)
1487 24.4 Collagen, NH3
1410 23.5 Amino acids, aspartic and glutamic acid
806 21.8 DNA: O-P-O symmetric stretching
847 20.8 Saccharides (α-glucose, maltose)
1645 20.2 Amide I (α-helix of Protein)
Results and Discussions
39
Table 4-2 Negative regression coefficient obtained by PLS model for NS1 based
screening along with molecular description.
Raman
shift
Strength of
regression
coefficients
(%)
Molecular assignments
1164 100.0 Tyrosine
1029 99.0 O-CH3 stretching of methoxy groups, keratin (protein
assignment)
780 74.0 Ring breathing of nucleotide bases
1230 71.0 DNA/RNA bases and amide III
1093 70.8 Phosphate backbone vibration as a marker mode for the
DNA concentration
1010 68.5 Tryptophan ring breathing
1055 60.6 RNA/DNA
1018 58.9 Ribose of RNA/DNA
760 56.2 Ethanolamine group, phosphatidyl-ethanolamine,
tryptophan (proteins)
1676 55.4 Amide-I (β-sheet) and DNA/RNA bases
1097 54.2 Phosphodioxy (PO2−) groups
1421 53.0 DNA/RNA bases and proteins
1294 51.6 Methylene twisting and amide III (protein band)
1666 47.3 α-Helical structure of amide-I (collagen assignment) and
DNA/RNA bases
1354 39.1 DNA/RNA bases
1067 38.0 PO2− stretching (DNA/RNA)
973 37.1 Ribose vibration, one of the distinct RNA modes
841 36.4 Saccharide (α)
858 36.4 DNA bases
1075 34.5 PO2− stretching (DNA/RNA)
1201 33.3 DNA/RNA
1117 33.0 Glucose
1682 32.9 One of absorption positions for the C=O stretching
vibrations of cortisone
1378 31.0 Paraffin, lipid assignment
1436 29.1 DNA,/RNA bases
883 27.8 Proteins, including collagen-I
1432 27.7 DNA/RNA and proteins
832 27.4 Asymmetric O-P-O stretching DNA bases
Results and Discussions
40
4.2 IgM based Screening
Diagnostic methods in practice are based on detection of DENV itself or its related
antibodies like IgM, which are produced as a response against DENV infection by the
human immune system, as discussed in Chapter 1 and Chapter 2 in details.
In the present study, a multivariate model has been developed to predict
quantitative values of AI of IgM in the dengue suspected samples. This model is
developed by utilizing the PLS regression. Raman spectra of 78 samples have been
used as the trainee data set. Fig. 4-4 displayed the patch area display and average of
pre-processed spectra of 78 samples used in the model development. ELISA based
cut-off value for AI of IgM was 9. Out of 78 DENV infected samples, 37 have IgM
values of AI above cut-off (≥ 9) (red colored) and 41 below cut-off (< 9) (green
colored). Testing data set contained 30 samples which were used for blind testing of
the developed model. Predicted values of AI of IgM by the model were found in
excellent agreement with the ELISA results.
Figure 4-4 Raman spectra of sera samples used in IgM based screening study.
4.2.1 Results and Discussions:
The maximum number of PCs to be used for model development was determined
according to the methods given in section 3.4. Firstly, the RMSE curve was obtained
Results and Discussions
41
as shown in Fig. 4-5. Secondly, methods of Kaiser, Scree and parallel factor [120] are
employed as shown in Fig. 4-6.
Figure 4-5 RMSECV curve calculated by using number of PCs from 1 to 20.
Figure 4-6 Curves obtained by employing methods of Kaiser, Scree and parallel
factor analysis for IgM based screening.
Results and Discussions
42
It produced three curves; the real data eigenvalues (blue color), percentile of
eigenvalues (red color) and mean of eigenvalues (green color). According to both of
these figures it is assured that the choice of 3 PCs is authenticated and acceptable to
avoid over-training of the model. The value of r2 for this model has been found to be
0.929. The second parameter is the RMSECV that is also used to validate the
outcomes of statistical model. In this model, RMSECV was calculated to be 2.17.
Based on LOO cross validation method, calibration curve has been plotted for IgM
using 78 samples, as shown in Fig. 4-7. The predicted values of AI of IgM are shown
in Figs. 4-7. Evidently, the predicted values of AI of IgM are quite promising and an
excellent correlation has been found with clinical results. The RMSEP of AI of IgM
for the blind samples has been found to be 3.25, which shows a reasonable accuracy
of the model.
Figure 4-7 Calibration curve for IgM based screening along with predictions for
testing data set.
Another important statistical parameter is SD in errors of the predicted values
used to evaluate the accuracy of multivariate model. Standard deviation in errors of
LOO predicted values of AI of IgM for 78 trainee samples were found to be 2.18,
whereas in 30 blind suspected samples, it was 3.31. Sensitivity, specificity and
accuracy have been calculated and plotted in Fig. 4-8 at the corresponding cut-off
Results and Discussions
43
values. Importantly, these parameters have been calculated from the predictions of 30
blindly tested samples. Accuracy, sensitivity and specificity were calculated to be
96.67 %, 90 % and 100 % respectively. An ROC curve, shown in Fig. 4-9, is also
produced and AUC was found to be 0.985.
Figure 4-8 Sensitivity, specificity and accuracy of the IgM based screening model
at different cut-off values.
Figure 4-9 ROC curve for IgM based screening.
Results and Discussions
44
Analysis of Regression Coefficients
To visualize the relevance of coefficients of regression (plotted as regression vector)
with spectral variations caused by IgM during infection, all the Raman spectra were
divided into three groups as shown in Fig. 4-10. Spectra with an AI value of IgM
below 9 are placed in group “IgM negative” and their average spectrum is plotted in
green color. Similarly, the spectra with an AI value of IgM between 9 and 20 are
placed in group “mild IgM positive” and their average spectrum is plotted in blue
color, while the spectra with AI value of IgM above 20 are placed in group “strong
IgM positive” and their average spectrum is plotted in red color. At the bottom of that
plot, regression curve is also displayed for comparison in black color. It was
established that regression curve has pointed out some Raman bands with positive and
negative regression coefficients. These trends were also confirmed by analyzing the
average spectra of these groups visually. Dengue virus infection causes the rising and
reducing of some molecular levels and inducing of new ones (IgM) which form the
basis of the regression vector of the model.
Figure 4-10 Regression vector along with average spectra of negative, mild IgM
positive and strong IgM positive samples.
Molecules associated with these Raman bands have been identified through
existing literature. Raman shift at 1594 cm-1 is assigned to asparagine [137] which has
been reported for its role in DENV propagation in the host [138]. Raman shifts at
1537, 1075 and 1318 cm-1 are assigned to glutamate which is reported to accumulate
Results and Discussions
45
due to DENV infection and its effects on the nervous system [139]. Raman shifts at
1514, 1387 and 1318 cm-1 are assigned to galactosamine which is reported to be an
integral part of IgM [140]. Raman shifts at 1387, 1437 and 1065 cm-1 are assigned to
palmitic acid which is reported for its high concentration due to IgM activation and
stress in the body [141] and It is also reported for its role in apoptosis in liver due to
DENV infection [142].
Raman shifts at 1387, 1018, 1075, 732 and 1359 cm-1 are assigned to dextrose
which is reported to be found in elevated levels in DENV infected subjects [143].
Raman shifts at 1018, 1437, 1507, 1473 and 1097 cm-1 are assigned to myristic acid
which is found in higher concentrations in serum due to IgM [141]. Raman shifts at
1018 and 1065 cm-1 are assigned to vaccenic acid which is converted to linoleic acid
[144] in the body and is reported for its role in the immune system [142]. Raman
shifts at 1437 and 1378 cm-1 are assigned to arginine which is reported to have high
concentration in DENV infection [145]. Raman shift at 1075 cm-1 is reported for
triglyceride which has been reported for its elevated level in serum because of IgM
[146]. Raman shifts at 1473, 1318 and 1097 cm-1 are assigned to
phosphoenolpyruvate which is reported to have high concentration in DENV infection
[147]. Raman shift at 732 cm-1 is assigned to phosphatidylserine, which is reported to
mediate the entry of DENV in target cells [148]. Raman shift at 626 cm-1 is assigned
to fructose which has shown a decreasing trend in the present study because it is
consumed to form fructose-bisphosphate-aldolase which takes part in glycolysis
which is necessary for DENV infection [149]. Raman shifts at 885 cm-1 are assigned
to cellobiose. Raman shifts at 936 and 843 cm-1 are assigned to arabianose. Both of
these moles have shown a decreasing trend when the IgM level rises in the body.
Raman shifts at 608 and 957 cm-1 are assigned to cholesterol which is reported for
their relation with IgM. Individuals, who have low cholesterol levels, produce higher
concentration of IgM [150]. A summary of these assignments is given in Table 4-3.
Results and Discussions
46
Table 4-3 Prominent Raman bands highlighted by the strongly positive or
strongly negative values of regression coefficients of IgM based model.
Raman
shift (cm-1)
Trend
with AI
of IgM
Assigned
molecule Information from literature
1594 positive Asparagine It plays an essential role in DENV
propagation.
1537,
1075, 1318 positive Glutamate
DENV infection induces its accumulation
by creating nervous disorder.
1514,
1387, 1318 positive Galactosamine It is a part of IgM.
1387,
1437, 1065 positive Palmitic acid
Its concentration increases with high level
of IgM in serum. It is also reported to
have a role in apoptosis in liver due to
DENV infection.
1387,
1018,
1075, 732,
1359
positive Dextrose It is found in elevated levels in DENV
infected subjects.
1018,
1437,
1507,
1473, 1097
positive Myristic acid Its concentration increases with high level
of IgM in serum.
1018, 1065 positive Vaccenic acid
Vaccinic acid is converted into linoleic
acid. Linleoic acid plays its role in
immune system.
1437, 1378 positive Arginine Its concentration rises due to DENV
infection.
1396,
1473,
1318, 1097
positive Phosphoenol-
pyruvate
Elevated levels of phosphoenolpyruvate
are reported in DENV infection.
1075, 1300 positive Triglycerides IgM levels show positive relationships
with triglycerides.
732 positive Phosphatidyl-
serine
It mediates the entry of DENV in target
cells.
626 negative Fructose
Due to DENV infection glycolysis takes
place, and it results in reduced levels of
fructose as it is consumed and its level
decreases.
608, 957 negative Cholesterol
Individuals with low cholesterol levels
exhibited higher IgM levels than
individuals without it.
885 negative Cellobiose Not found in already published literature,
having relation with DENV infection
936, 843 negative Arabinose Not found in already published literature,
having relation with DENV infection
Results and Discussions
47
4.3 IgG based Screening
In the present study, a multivariate model was developed to predict quantitative
values of AI of IgG in the dengue suspected samples. This model is developed by
utilizing the PLS regression, using functions in the MATLAB (Mathworks 2009a).
For training of the model, 79 sera samples have been used. Clinically approved
method, ELISA, has been used for the determination of AI of IgG. These values of AI
of IgG have been provided to the multivariate model for training. For the validation of
model, a blind test was performed using 20 unknown suspected samples. The clinical
results of ELISA for these samples were kept hidden during model development.
Spectral range was chosen from 500-1600 cm-1. Fig. 4-11 displayed patch area display
and average of pre-processed spectra of 79 samples used in the model development.
ELISA based cut-off value for AI of IgG was 9. Out of 79 DENV suspected samples,
36 have IgG values of AI above cut-off (≥ 9) (red colored) and 43 below cut-off (< 9)
(green colored). In addition, Raman spectra of 20 suspected samples were kept hidden
for blind evaluation of model. Optimization of the number of PCs to be used for
model development is done by two methods. According to Fig. 4-12 and Fig. 4-13 it
was concluded that the choice of 3 PCs is authenticated and acceptable to avoid over-
training of the model.
4.3.1 Results and Discussions
Leave one sample out cross validation method has been applied for calibration.
During the LOO process, two parameters have been calculated to check the goodness
of the model; one of these is r2, which explains the variability level of outcomes in the
model and gives a guideline to measure the validity of the model. The value of r2 for
this model has been found to be 0.91, whereas the value of r2 greater than 0.9 is
accepted clinically [121], [151]. In this model, RMSECV for the entire LOO cross
validated predicted values were calculated to be 2.06. It showed the robustness and
authentication of the multivariate model for the prediction of AI of IgG in the samples
[152]. When a suspected Raman spectrum is loaded by GUI platform, it is multiplied
by the corresponding regression vector and produces a predicted value for AI of IgG.
Based on LOO cross validation method, calibration curve has been plotted for IgG
using 79 samples, as shown in Fig. 4-14. The prediction of AI of IgG in 20 suspected
samples, whose clinical ELISA results were kept hidden, are also shown in the same
Results and Discussions
48
figure. These predicted values were quite promising and an excellent correlation has
been found with clinical results. The RMSEP of AI of IgG for these suspected blind-
tested samples has been found to be 3.25, which shows a reasonable accuracy of the
model [121], [151].
Figure 4-11 Patch area display of Raman spectra used in IgG based screening.
Figure 4-12 RMSE curve for PCs optimization for IgG based model.
Results and Discussions
49
Figure 4-13 Eigen values based curves for optimization for IgG based model.
Standard deviation of errors in the predicted values used to evaluate the
accuracy of multivariate model. Standard deviation of errors in LOO cross-validation
predicted values of AI of IgG for 79 trainee samples were found to be 2.18, whereas
in 20 blind suspected samples, it was 3.31. These predicted values were awarded zero
point score because these were within one standard deviation [152], and according to
the scoring system of QCMD [153], these results qualify for clinical acceptance.
Sensitivity, specificity, accuracy and area are under ROC curve are very importance
parameters to assess the goodness of fit for a statistical model for medical application
[51]. From the predictions of 20 blindly tested samples; sensitivity, specificity and
accuracy at cut-off have been calculated to be 100 %, 83.3 % and 95 % respectively
as shown in Fig. 4-15. Receiver operating characteristic curve, shown in Fig. 4-16, is
produced by plotting the true-positive rate against false-positive at various cut-off
values from −0.36 to 27.7. Area under ROC curve was found to be approximately 1,
which indicates the reasonable level of accuracy of this model [125].
Results and Discussions
50
Figure 4-14 Calibration curve of PLS model developed for IgG based screening.
Figure 4-15 Sensitivity, specificity and accuracy of the PLS model for IgG at
different cut-off values.
Results and Discussions
51
Figure 4-16 Receiver operator characteristic (ROC) curve for IgG based
screening model.
A vector consisting of regression coefficients is yielded by PLS regression. It
is called regression vector. Relevance of coefficients of regression (plotted as
regression vector) with spectral variations caused by IgG during infection is studied
by dividing all the Raman spectra into three groups as shown in Fig. 4-17. Spectra
with AI value below 9 are placed in group “IgG negative” and their average spectrum
is plotted in green color. Similarly, the spectra with AI value between 9 and 20 are
placed in group “mild IgG positive” and their average spectrum is plotted in blue
color, while the spectra with AI value above 20 are placed in group “strong IgG
positive” and their average spectrum is plotted in red color. At the bottom of that plot,
regression curve is also displayed for comparison in black color. It was established
that regression curve has pointed out some Raman bands with positive and negative
trends which are confirmed by analyzing the average spectra of these groups visually.
The Raman bands where regression coefficients are positive indicate those molecules
whose concentrations are increasing with rising levels of biochemical changes
associated with the AI of IgG, whereas, the Raman bands where regression
coefficients are negative indicate the molecules whose concentrations are decreasing
Results and Discussions
52
with increasing value of AI of IgG. Molecules associated with these Raman bands
have been identified through existing literature [126], [127] and are listed in Table 4-4
and Table 4-5.
Figure 4-17 Regression vector along with average spectra of negative, mild IgG
positive and strong IgG positive samples.
Raman shift at 1245 1272 1287 and 1262 cm-1 are assigned to amide III.
Raman shifts at 873 1330 1454 and 934 cm-1 are assigned to collagen and Raman shift
at 1454 1443 1454 1466 1477 1443 1466 and 1443 cm-1 are assigned to proteins.
These three interrelated biomolecules are found to be positively correlated with AI
value of IgG. Raman shift at 1443 cm-1 is assigned to Fatty acids and Raman shift at
1454 cm-1 is assigned to Phospholipids. Lipid profiling is a topic of interest for the
diagnosis of DENV infection [154]. In the present study, these lipids/fatty acids are
found to be negatively correlated with an increased value of AI of IgG. Raman shift at
529 cm-1 is assigned to Fucose [126] that might be involved in the process of
fucosylation during dengue infection as dengue E protein has two potential
glycosylation sites [155]. Raman shifts at 982, 1018, 1063, 1099 and 1379 cm-1 are
assigned to myristic acid. It was reported to be positively correlated with IgM [51]
and the present study has shown a similar trend with IgG also. Raman shifts at 1095,
Results and Discussions
53
1575, 633, 721, 1099 and 651 cm-1 are assigned to coenzyme-A. It is required by
DENV to increase the cellular fatty acid synthesis [156]. Raman shifts at 603, 633,
1043, 1055, 1095 and 1191 cm-1 are assigned to glutamate. It is reported that DENV
infection induces glutamate excitotoxicity by releasing more glutamate in synaptic
cleft [157]. In the present study, it is found to be positively correlated with AI of IgG
and it was reported to be positively correlated in IgM study as well. Raman shifts at
1419 cm-1 and 1379 cm-1 are assigned to Alanine. Glutamate is converted into alanine,
which is released into the bloodstream [158]. Raman shifts at 1079 cm-1 and 597 cm-1
are assigned to amide II and amide VI respectively. These types of amides are found
to be positively correlated with IgG in the present study. Raman shifts at 1137 cm-1
and 1095 cm-1 are assigned to arabinose. In IgM study, it was negatively correlated
while in the present study it is found to be positively correlated. The exact role of
arabinose is yet to be discovered, however, it is observed that abdominal pain is a
symptom of DENV infection and arabinose is biomarker of abdominal pain due to
yeast infection [159] and arabinose is used as a culturing medium for viruses [160].
Raman shifts at 1099 cm-1 and 982 cm-1 are assigned to arginine. Its concentration
rises due to DENV infection [145]. Raman shifts at 567 cm-1 and 588 cm-1 are
assigned to vitamin C (ascorbic acid). It is positively correlated with AI of IgG.
Vitamin C plays a role in the immune system [161], [162] and concentration of IgG
rises with vitamin C [163]. Raman shifts at 1191, 1119, 1150, 1154 cm-1 are assigned
to carotene It is found in the present study that carotene is positively correlated with
IgG. Raman shifts at 982 cm-1 and 1594 cm-1 are assigned to fumarate It is related to
immune modulation [164]. Raman shifts at 703, 1063, 1095, 1154 and 1587 cm-1 are
assigned to galactosamine. Its concentration rises due to IgM [51]. This study reveals
that it is also related to IgG as it is found to be positively correlated with IgG. Raman
shifts at 749 cm-1 are assigned to lactic acid. It is reported that a high level of lactic
acid is a potential biomarker for diagnosis of DENV infection [53]. Raman shifts at
1063, 1099, 1419 cm-1 are assigned to Stearic acid. It is up-regulated due to DENV
infection [165]. Raman shifts at 597, 765, 1119, 1575, 1079, 574, 749 and 1363 cm-1
are assigned to tryptophan. It is reported to be elevated during DENV infection [23].
Raman shifts at 1018 and 1079 cm-1 are assigned to vaccenic acid. It is found to be
positively correlated with IgG in the present study. It was also reported in IgM study
that vaccenic acid is converted into linoleic acid. Linoleic acid plays its role in the
immune system. Raman shifts at 1587, 621 and 1182 and 1002 cm-1 are assigned to
Results and Discussions
54
phenylalanine. Interestingly, it was found that at 1002 cm-1 a strong lowering trend is
shown while other bands of phenylalanine at 1587, 621 and 1182 cm-1 were strongly
positive. This needs to be investigated further in detail. However, it is reported in a
study that its concentration is found to be higher in DENV infected subjects [166].
Normally, phenylalanine is converted into tyrosine by kidney and liver, but due to
malfunction of these organs this conversion is stopped and level of phenylalanine
rises [167].
Table 4-4 Prominent Raman bands which have been highlighted by the strongly
negative values of regression coefficients of this model are tabulated for their
bio-molecular assignment.
Raman
Shift
(cm-1)
Strength
of
Regression
coefficient
(%)
Bio-Molecule Molecular description
1245 100
Amide III
These bands of proteins are found to be
negatively correlated in the present study,
specifically collagen and amide-III.
1272 72
1287 53
1262 61
873 30
Collagen 1330 24
1454 69
934 55
1443 54
Proteins 1466 33
1477 28
1466 33
1443 54 Fatty acids Lipid profiling is a topic of interest for the
diagnosis of DENV infection. 1454 69 Phospholipids
529 28
Fucose Fucose is involved in the process of
fucosylation during dengue infection. 1272 72
1330 24
Results and Discussions
55
Table 4-5 Prominent Raman bands which have been highlighted by the strongly
positive values of regression coefficients of this model are tabulated for their bio-
molecular assignment.
Raman
Shift
(cm-1)
Strength of
Regression
coefficient
(%)
Bio-Molecule Molecular description
982 59
Myristic acid IgM study has shown similar trend also.
1018 52
1063 47
1099 33
1379 64
1095 34
Coenzyme-A It is required by DENV to increase the
cellular fatty acid synthesis.
1575 32
633 86
721 44
1099 33
1419 23
651 67 Alanine
It is mainly related to liver damage which
occurs in dengue infection. 1379 64
1079 34 Amide II
Amide VI
These types of amides are found to be
positively correlated with IgG in the
present study. 597 70
1095 34 Arabinose
It is observed that abdominal pain is a
symptom of DENV infection and
arabinose is biomarker of abdominal pain. 1137 87
982 59 Arginine
Its concentration rises due to DENV
infection [145]. 1099 33
567 38 Vitamin C
Vitamin C plays a role of antioxidant in
DENV infection [162]. 588 66
1191 88
Carotene It is positively correlated with IgG.
1119 38
1150 76
1154 77
982 59 Fumarate It has a role in immune modulation [164].
1594 42
703 39
Galactosamine It was earlier reported to be high in IgM
study as well [51].
1063 47
1095 34
1154 77
1587 42
603 59
Glutamate It is reported that DENV infection
induces glutamate excitotoxicity.
633 86
1043 32
1055 49
1095 34
1191 88
Results and Discussions
56
749 43 Lactic acid
It is reported that a high level of lactic
acid is a potential biomarker for diagnosis
of DENV infection [53].
1587 42
Phenylalanine
Interestingly. Raman bands of
phenylalanine are positively correlated
except 1002 cm-1. It needs to be further
investigated.
621 82
1182 100
1002 -52*
1063 47
Stearic acid It is up-regulated due to DENV infection
[165]. 1099 33
1419 23
597 70
Tryptophan It is reported to be elevated during
DENV infection [23].
765 38
1119 38
1575 32
1079 34
574 42
749 43
1363 54
1018 52 Vaccenic acid
Vaccenic acid is converted into linoleic
acid. Linoleic acid plays its role in the
immune system. 1079 34
* This specific Raman band of phenylalanine is found to be negatively correlated with
IgG.
Results and Discussions
57
4.4 Lactate as Biomarker
In present study, the diagnosis of DENV infection in human blood sera based on an
increase in lactate concentration using Raman spectroscopy is investigated. For
strengthening the claim, different concentrations of lactic acid solution has been
added to normal/healthy sera. The changes occurred in the composition of the serum
sample have been analyzed and discussed accordingly. This will be quite helpful in
the early diagnosis of DENV infection which is of prime importance for management
of the disease.
In total 70 samples of different ages and genders have been used in the present
study. Among these, 20 samples were from healthy volunteers whereas, 50 were
obtained from DENV infected patients. In one part of healthy sera two different
concentration 50 mM/L and 100 mM/L of lactic acid solution (L 1250, Sigma-Aldrich
Chemie GmbH, Germany) were prepared in a control manner for observing their
effects. The sample collection, preparation and storage procedure is same as
mentioned in section 3.1.
4.4.1 Acquiring Raman Spectra
Raman spectra from DENV infected sera, healthy sera, lactic acid solution and lactic
acid solution mixed with healthy sera were recorded. About 15 µl of each sample has
been put on the glass slide and left for some time at room temperature for water
moisture to vaporize. The schematic diagram of the experimental setup is shown in
Fig. 4-18. Raman spectrometer (µRamboss DONGWOO OPRTON, South Korea)
with a spectral resolution of 4 cm-1 was used for recording Raman spectra from all
samples. A diode laser emitting @532 nm has been used for the excitation. The
measured laser power at the sample surface was 40 mW. A microscope objective
having a magnification of 100X has been used, both for focusing the light on the
sample and collection of backscattering light. An acquisition time of 10 seconds has
been used for recording each spectrum. A spectral range from 600 to 1800 cm-1 has
been selected for recording Raman spectra.
Results and Discussions
58
Figure 4-18 Sketch of experiment setup.
4.4.2 Raman spectral analysis
All the Raman spectra have been smoothed using ‘Savitzky-Golay’ [111] filter with
five points and 3rd order polynomial fitting. The mean vector normalized Raman
spectra of healthy and dengue infected sera as well as the mean difference between
normal and infected samples are shown in Fig. 4-19. In healthy sera three intense
Raman peaks appeared at 1003, 1156 and 1516 cm-1. In dengue infected samples,
Raman peaks appeared at 750, 830, 925, 950, 1003, 1123, 1156, 1333, 1450, 1516,
1580, 1680 and 1730 cm-1.
Figure 4-19 Vector normalized mean Raman spectra of healthy and dengue
infected sera (upper) along with the mean difference between the normal and
infected samples (lower).
Results and Discussions
59
Furthermore, obvious differences between the normal and dengue infected
samples appeared at 750, 830, 925, 950, 1003, 1123, 1156, 1450, 1516, 1580, and
1730 cm-1 as can be seen in the difference plot in blue color. The detailed assignment
of most of these Raman peaks has been given in another article [22]. Fig. 4-20
illustrated the vector normalized Raman spectra of lactic acid solution with two
intense Raman peaks at 830 cm-1 and 1450 cm-1 along with some medium intensity
peaks at 750, 875, 925, 1040, 1075 and 1730 cm-1.
Figure 4-20 Vector normalized Raman spectra of lactic acid solution.
Fig. 4-21 illustrated the recorded vector normalized mean Raman spectra of healthy
blood sera; dengue infected sera as well as two different concentration of lactic acid
solution mixed with healthy sera. For an obvious differentiation, Raman spectra of
healthy sera samples are shown in green color, dengue infected in red color, whereas
lactic acid solution mixed with healthy sera are shown in blue color (50 mM/L) and
magenta color (100 mM/L).
Results and Discussions
60
Figure 4-21 Vector normalized mean Raman spectra of healthy sera, dengue
infected sera, 50 mM/L and 100 mM/L of lactic acid solution in healthy sera.
4.4.3 Results and Discussion
The Raman peaks appeared in normal human blood sera have been explained
previously [22], [23]. Raman peak at 1003 cm-1 has been assigned to symmetric ring
breathing mode of phenylalanine and β-carotene, whereas the peaks at 1156 and 1516
cm-1 have been assigned to β-carotene [18], [168], [169]. These Raman peaks are
highly reproducible. In DENV infected sera, these three peaks are either suppressed or
their intensity is decreased. Moreover, new peaks were also arisen at different
frequencies. In dengue infected sera, additional Raman peaks appeared at 750, 830,
925, 950, 1123, 1333, 1450, 1580, 1680 and 1730 cm-1 as shown in Fig. 4-19. The
main contribution to these spectral lines are most probably corresponds to a high
concentration of lactate in DENV infected sera. In the human body, lactate is
produced continuously mostly in muscles. It is then transported to different metabolic
organs via blood which regulates them [170]–[172]. Liver is considered to be the key
organ that converts blood lactate into pyruvate. Around 50-70 % of blood lactate is
extracted by liver and converted into pyruvate [168]. An additional amount of lactate
is cleared by the kidney and some other organs. In normal conditions, with adequate
tissue perfusion, conversion of pyruvate to Acetyl-CoA is occurred largely bypassing
lactate production. In tissue hypoxia/hypo-perfusion, lactate is produced as an end
product of pyruvate in the presence of lactate dehydrogenase enzyme. Lactate exists
Results and Discussions
61
in two isoforms, L-and D-lactate, such that L-lactate is the primary isomer produced
in the human body. The biochemical impact of DENV infection on the function of
various body organs like liver, kidney, lungs, heart etc. as well as an elevated level
of lactate is well established and reported [173]–[179]. As stated earlier, liver and
kidney are the two main organs in the human body which regulate the lactate level.
Changes in hepatic oxygen supply and intrinsic hepatic disorder affect the
capacity of the liver to metabolize lactate. In such condition, liver becomes a lactate
producing organ rather than using it for gluconeogenesis. Hence, due to dysfunction
of these important body organs in dengue infection, blood lactate level increases. A
good agreement has been observed by comparing the Raman peaks of lactic acid
solution (Fig. 4-20) and dengue infected sera samples (Fig. 4-19). More precisely,
Raman peaks close to 750, 830, 925, 1123, 1450 and 1730 cm-1 appeared both in
lactic acid solution as well as DENV infected samples. Furthermore, a slight blue shift
at wave number 1003 cm-1occurs in dengue infected samples. Possible cause this shift
in dengue is not clear, but it could be due to somewhat different protein composition
next to phenylalanine and β-carotene. For the observation of lactate effects on normal
blood sera, two different concentration of lactic acid solution (50 mM/L and 100
mM/L) has been added to healthy sera in a controlled manner and their Raman spectra
have been recorded as shown in Fig. 4-21. A gradual decrease in the intensity of the
aforementioned three peaks has been observed with an increase in the concentration
of lactic acid solution in the sera samples is clearly visible. So, it can be said that,
apart from the carotenoids deficiency in DENV infection as described earlier [23], the
suppression of Raman peaks at 1003, 1156, and 1516 cm-1 in healthy sera may also be
attributed to an elevated lactate level in the blood. One can use the appearance of
lactate in blood sera as a valuable indicator for the presence of disease. In order to
evaluate lactate as a potential biomarker for dengue diagnosis, in depth studies will be
necessary.
In conclusion, this study presents the screening of DENV infection in human
blood sera based on lactate concentration using Raman spectroscopy. A total of 70
samples, 50 from confirmed DENV infected patients and 20 from healthy volunteers
have been used in the present study. Raman spectra of all these samples have been
acquired in the spectral range from 600 cm-1 to 1800 cm-1 using 532 nm laser as an
excitation source. Spectra of all these samples have been analyzed for assessing the
Results and Discussions
62
biochemical changes resulting from infection. In DENV infected samples three
prominent Raman peaks have been found at 750, 830 and 1450 cm-1. These peaks are
most probably attributed to an elevated level of lactate due to an impaired function of
different body organs in dengue infected patients. This has been proven by an addition
of lactic acid solution to the healthy serum in a controlled manner. By the addition of
lactic acid solution, the intense Raman bands at 1003, 1156 and 1516 cm-1 found in
the spectrum of healthy serum got suppressed, while new peaks appeared at 750, 830,
925, 950, 1123, 1333, 1450, 1580 and 1730 cm-1. The current study predicts that
lactate may possibly be a potential biomarker for the diagnosis of DENV infection.
63
5 Conclusions and Future Prospects
Raman spectroscopy has great potential as a technique to be used for the diagnosis of
DENV infection. The present research work has shown that it can be used
qualitatively as well as quantitatively for the screening purpose. The dengue virus
produces NS1 and the response of human body results in the production of IgM and
IgG. Raman spectroscopy has successfully been used to screen the sera samples of
subjects, suspected of DENV infection, on the basis of NS1, IgM and IgG. Moreover,
an elevated level of lactate is also shown to be an important biomarker for the
diagnosis of DENV infection.
It is important to mention here that the symptoms of DENV infection are
somehow similar to the other diseases like flu, typhoid, malaria, pneumonia, measles,
enteric fever, leptospirosis, typhus fever [180] etc. Differentiating a healthy group
from an infected group is usually easier because of so many molecular changes that
occur in the blood due to infections of different kinds. However a physician usually
deals with the subjects who are suspected of DENV infection based on the symptoms
which are quite similar to other diseases. Usually, such subjects have to give samples
to be tested for all of these diseases separately to accurately diagnose DENV infected
subjects. In these studies we have selected the suspected subjects which were
presented in the hospital with symptoms similar to the DENV infection. These
subjects may have been suffering from malaria, typhoid, flu etc. Raman spectra of
clinically confirmed DENV infected subjects were used to train the PLS regression
models, which successfully differentiated the DENV infected subjects and the non-
DENV infected samples. This study confirms that the proposed technique has shown
potentials to differentiate DENV infected sample from all the suspected samples.
Conclusions and Future Prospects
64
5.1 Raman Spectroscopy based Diagnosis of DENV
Infection
5.1.1 NS1 based Study
In the present study, a multivariate model has been developed to differentiate NS1
positive and NS1 negative samples in the DENV suspected subjects. Analysis of
DENV suspected samples highlights phosphatidylinositol, ceramide, proline and
thiocyanate at 776, 1127, 1454 and 736 cm-1, respectively. These molecules have been
reported in literature for their role in DENV infection. These molecules have been
identified as potential biomarkers of DENV infection in the present study. Further
research work on these molecules relevant to NS1 may help to develop Raman
spectroscopy as efficient, reliable, and cost effective diagnostic tool for the
recognition of early DENV infection through the prediction of NS1 positive or NS1
negative samples.
5.1.2 IgM based Study
Raman spectroscopy based PLS regression model has been developed using 78
Raman spectra of DENV suspected sera for the prediction of AI of IgM. The model
was operated through GUI platform that loads the Raman spectrum of the suspected
sample. It is multiplied with regression vector of the model and predicts AI of IgM.
The predicted values of AI of IgM are in good agreement with clinical results.
Analysis of regression coefficients revealed that asparagine, glutamate,
galactosamine, palmitic acid, dextrose, myristic acid, vaccenic acid, triglycerides,
phosphoenolpyruvate and phosphatidylserine were found to have an increasing trend
with increasing values of AI of IgM. However, fructose, cholesterol, cellobiose and
arabianose were found to have a decreasing trend with an increasing value of AI of
IgM. The model predicted values were based on reference clinical results; however,
further research is in progress at NILOP to investigate in details for finding the pin-
pointed Raman signatures in the infected samples for IgM.
Conclusions and Future Prospects
65
5.1.3 IgG based Study
Raman spectroscopy based PLS regression model has been developed using 79
Raman spectra of DENV suspected sera for the prediction of AI of IgG. The model
was operated through GUI platform that loads Raman spectrum of the suspected
sample, multiplies it with regression vector of the model and predicts AI of IgG. The
predicted values of AI of IgG are in good agreement with the clinical results.
Molecular analysis on the basis of regression coefficients revealed that myristic acid,
coenzyme-A, alanine, arabinose, arginine, vitamin C, carotene, fumarate,
galactosamine, glutamate, lactic acid, stearic acid, tryptophan and vaccenic acid are
positively correlated with the values of AI of IgG. However, amide III, collagen,
proteins, fatty acids, phospholipids and fucose are negatively correlated with values of
AI of IgG. The model predicted values were based on reference clinical results;
however, further research is in progress at NILOP to investigate in details about
finding the pin-pointed Raman signatures in the infected samples for IgG.
5.1.4 Lactate as a Biomarker
Lactate based detection of DENV infection in human blood sera using Raman
spectroscopy was investigated. In dengue infected samples, Raman peaks appeared at
750, 830, 925, 950, 1123, 1333, 1450, 1580, 1680 and 1730 cm-1. In dengue infected
samples, the Raman peaks close to 750, 830, 925, 950, 1123 and 1450 cm-1 are most
probably showing an elevated lactate level which arises due to the impaired function
of important body organs like liver, kidney, lungs etc. Furthermore, it has also been
shown that in DENV infection, the Raman peaks at 1003, 1156, and 1516 cm-1
suppress most probably due to an elevated lactate level in the blood. The research
work at NILOP is still in progress and efforts are underway to provide an alternate
and efficient tool that might help in early detection of different diseases.
5.2 Future Prospective
Cost of Raman system used in presented research work is high as it provides various
options as per requirement of the experiments. However, acquisition of Raman
spectrum does not cost much and results are very quick.
Conclusions and Future Prospects
66
Early diagnosis of DENV infection is of prime importance because a DENV
infected patient, who is diagnosed at an early stage, can be treated well in time and its
symptoms are administrated in an efficient way so that the infection may not lead to
DHF and DSS. In this regard a study was conducted [50], where sera samples of
subjects with possible DENV infection were collected and their ELISA based clinical
reports were acquired about NS1, IgM and IgG. The subjects for whom all three tests
were negative were placed in control group while the positive group consisted of only
those samples for which NS1 was positive but IgM and IgG were negative. Hence we
had two groups; one was non-DENV infected and the other with DENV infection but
at an early stage. To differentiate these two groups a PLS model was successfully
implemented with very good results. It showed that the Raman spectroscopy based
PLS model can diagnose DENV infection at an early stage when only NS1 are present
in the serum and antibodies have not yet been produced. Embedding of such a code in
a microcontroller with integrated hand held Raman systems can help diagnose the
DENV infection at an early stage as well.
The GUI used for this research work was designed and developed specifically
in MATLAB (Mathworks 2009a) programming language to yield screening models,
i.e. regression vectors, for screening of serum samples. Such models have potentials
to be used in the code of a microcontroller in a Raman spectrometer which will enable
us to screen the sera samples instantly for NS1, IgM and IgG just like a glucometer,
which is commonly being used for glucose level determination. Such devices would
be able to acquire Raman spectrum of serum of a subject who is suspected of DENV
infection. It will classify a sample as positive or negative on the basis of NS1, IgM
and IgG. Moreover, elevated level of lactate in the serum may also be examined at the
same time. In this way it may help a physician in diagnosis of DENV infected
subjects accurately, quickly and cost effectively. Sooner or later this technique will be
approved for clinical trial. After successful testing of this technique in various
hospitals it can be developed into a hand held diagnostic device just like a glucometer.
67
References
[1] R. Perera and R. J. Kuhn, “Structural proteomics of dengue virus,” Curr. Opin.
Microbiol., vol. 11, no. 4, pp. 369–377, Aug. 2008.
[2] O. Horstick, Y. Tozan, and A. Wilder-Smith, “Reviewing Dengue: Still a
Neglected Tropical Disease?,” PLoS Negl. Trop. Dis., vol. 9, no. 4, p.
e0003632, Apr. 2015.
[3] S. Bhatt, P. W. Gething, O. J. Brady, J. P. Messina, A. W. Farlow, C. L.
Moyes, J. M. Drake, J. S. Brownstein, A. G. Hoen, O. Sankoh, M. F. Myers, D.
B. George, T. Jaenisch, G. R. W. Wint, C. P. Simmons, T. W. Scott, J. J.
Farrar, and S. I. Hay, “The global distribution and burden of dengue,” Nature,
vol. 496, no. 7446, pp. 504–507, Apr. 2013.
[4] D. A. Muller, A. C. I. Depelsenaire, and P. R. Young, “Clinical and Laboratory
Diagnosis of Dengue Virus Infection,” J. Infect. Dis., vol. 215, no. suppl_2, pp.
S89–S95, 2017.
[5] N. Gupta, S. Srivastava, A. Jain, and U. C. Chaturvedi, “Dengue in India,”
Indian J. Med. Res., vol. 136, no. 3, pp. 373–390, Nov. 2012.
[6] C. E. G. Smith, “The history of dengue in tropical Asia and its probable
relationship to the mosquito Aedes aegypti,” Trop. Med. Hyg., vol. 59, no. 10,
pp. 243–251, Oct. 1956.
[7] S. B. Halstead, S. Nimmannitya, and S. N. Cohen, “Observations related to
pathogenesis of dengue hemorrhagic fever. IV. Relation of disease severity to
antibody response and virus recovered.,” Yale J. Biol. Med., vol. 42, no. 5, pp.
311–328, Jan. 1970.
[8] R. J. Kuhn, W. Zhang, M. G. Rossmann, S. V. Pletnev, J. Corver, E. Lenches,
C. T. Jones, S. Mukhopadhyay, P. R. Chipman, E. G. Strauss, T. S. Baker, and
J. H. Strauss, “Structure of dengue virus: Implications for flavivirus
organization, maturation, and fusion,” Cell, vol. 108, no. 5, pp. 717–725, Mar.
2002.
[9] K.-I. Yamada, T. Takasaki, M. Nawa, and I. Kurane, “Virus isolation as one of
the diagnostic methods for dengue virus infection,” J. Clin. Virol., vol. 24, no.
3, pp. 203–209, Apr. 2002.
[10] R. G. Jarman, A. Nisalak, K. B. Anderson, C. Klungthong, B.
Thaisomboonsuk, W. Kaneechit, S. Kalayanarooj, and R. V. Gibbons, “Factors
influencing dengue virus isolation by C6/36 cell culture and mosquito
inoculation of nested PCR-positive clinical samples,” Am. J. Trop. Med. Hyg.,
vol. 84, no. 2, pp. 218–223, Feb. 2011.
[11] L. Rosen and D. Gubler, “The use of mosquitoes to detect propagate dengue
68
viruses,” Am J Trop Med Hyg, vol. 23, no. 6, pp. 1153–1160, Nov. 1974.
[12] Y. R. Lee, K. J. Huang, H. Y. Lei, S. H. Chen, Y. S. Lin, T. M. Yeh, and H. S.
Liu, “Suckling mice were used to detect infectious dengue-2 viruses by
intracerebral injection of the full-length RNA transcript,” Intervirology, vol.
48, no. 2–3, pp. 161–166, Mar. 2005.
[13] T. M. Yuill, P. Sukhavachana, A. Nisalak, and P. K. Russell, “Dengue-virus
recovery by direct and delayed plaques in LLC-MK2 cells.,” Am. J. Trop. Med.
Hyg., vol. 17, no. 3, pp. 441–448, May 1968.
[14] R. S. Lanciotti, C. H. Calisher, D. J. Gubler, G. J. Chang, and A. V Vorndam,
“Rapid detection and typing of dengue viruses from clinical samples by using
reverse transcriptase-polymerase chain reaction.,” J. Clin. Microbiol., vol. 30,
no. 3, pp. 545–551, Mar. 1992.
[15] B. Raengsakulrach, A. Nisalak, N. Maneekarn, P. T. Yenchitsomanus, C.
Limsomwong, A. Jairungsri, V. Thirawuth, S. Green, S. Kalayanarooj, S.
Suntayakorn, N. Sittisombut, P. Malasit, and D. Vaughn, “Comparison of four
reverse transcription-polymerase chain reaction procedures for the detection of
dengue virus in clinical specimens,” J Virol Methods, vol. 105, no. 2, pp. 219–
232, Sep. 2002.
[16] K. B. Chua, B. Mustafa, a H. Abdul Wahab, Y. K. Chem, a H. Khairul, V.
Kumarasamy, M. Mariam, H. Nurhasmimi, and K. Abdul Rasid, “A
comparative evaluation of dengue diagnostic tests based on single- acute serum
samples for laboratory confirmation of acute dengue,” Malays. J. Pathol., vol.
33, no. 1, pp. 13–20, 2011.
[17] S. Alcon, A. Talarmin, M. Debruyne, A. Falconar, V. Deubel, and M. Flamand,
“Enzyme-linked immunosorbent assay specific to dengue virus type 1
nonstructural protein NS1 reveals circulation of the antigen in the blood during
the acute phase of disease in patients experiencing primary or secondary
infections,” J. Clin. Microbiol., vol. 40, no. 2, pp. 376–381, Feb. 2002.
[18] M. Bilal, M. Saleem, S. T. Amanat, H. A. Shakoor, R. Rashid, A. Mahmood,
and M. Ahmed, “Optical diagnosis of malaria infection in human plasma using
Raman spectroscopy,” J. Biomed. Opt., vol. 20, no. 1, p. 17002, 2015.
[19] M. Bilal, M. Bilal, M. Saleem, M. Khurram, S. Khan, R. Ullah, H. Ali, M.
Ahmed, S. Shahzada, and E. U. Khan, “Raman spectroscopy based
investigation of molecular changes associated with an early stage of dengue
virus infection,” Laser Phys., vol. 27, no. 4, 2017.
[20] S. Khan, R. Ullah, A. Khan, A. Sohail, N. Wahab, M. Bilal, and M. Ahmed,
“Random Forest-Based Evaluation of Raman Spectroscopy for Dengue Fever
Analysis,” Appl. Spectrosc., vol. 71, no. 9, p. 370281769557, Mar. 2017.
[21] S. Khan, R. Ullah, A. Khan, N. Wahab, M. Bilal, and M. Ahmed, “Analysis of
dengue infection based on Raman spectroscopy and support vector machine
(SVM),” Biomed. Opt. Express, vol. 7, no. 6, pp. 2249–2256, 2016.
[22] S. Khan, R. Ullah, M. Saleem, M. Bilal, R. Rashid, I. Khan, A. Mahmood, and
69
M. Nawaz, “Raman spectroscopic analysis of dengue virus infection in human
blood sera,” Opt. - Int. J. Light Electron Opt., vol. 127, no. 4, pp. 2086–2088,
2016.
[23] M. Saleem, M. Bilal, S. Anwar, A. Rehman, and M. Ahmed, “Optical diagnosis
of dengue virus infection in human blood serum using Raman spectroscopy,”
Laser Phys. Lett., vol. 10, no. 3, p. 35602, 2013.
[24] M. Bilal, M. Bilal, M. Saleem, S. Khan, R. Ullah, K. Fatima, M. Ahmed, A.
Hayat, S. Shahzada, and E. U. Khan, “Raman spectroscopy based screening of
hepatitis C and associated molecular changes,” Laser Phys. Lett., vol. 14, no. 9,
2017.
[25] S. Ahmad, R. Ahmad, M. Bilal, and N. U. Rehman, “DFT studies of
thermoelectric properties of R-Au intermetallics at 300 K temperature,” J. Rare
Earths, 2017.
[26] S. A. Abbasi, A. H. Dogar, B. Ilyas, S. Ullah, M. Rafique, and A. Qayyum,
“Ion charge state and energy enhancement by axial magnetic field applied
during laser produced plasma expansion,” Laser Part. Beams, vol. 34, no. 4,
pp. 606–614, Dec. 2016.
[27] R. S. Das and Y. K. Agrawal, “Raman spectroscopy: Recent advancements,
techniques and applications,” Vib. Spectrosc., vol. 57, no. 2, pp. 163–176, Nov.
2011.
[28] A. J. Berger, T.-W. Koo, I. Itzkan, G. Horowitz, and M. S. Feld,
“Multicomponent blood analysis by near-infrared Raman spectroscopy,” Appl.
Opt., vol. 38, no. 13, p. 2916, 1999.
[29] A. M. K. Enejder, T.-W. Koo, J. Oh, M. Hunter, S. Sasic, M. S. Feld, and G. L.
Horowitz, “Blood analysis by Raman spectroscopy,” Opt. Lett., vol. 27, no. 22,
pp. 2004–2006, Nov. 2002.
[30] A. Fatima and J. Wang, “Review: progress in the diagnosis of dengue virus
infections and importance of point of care test: a review,” Pak. J. Pharm. Sci.,
vol. 28, no. 1, pp. 271–280, 2015.
[31] R. W. Peeling, H. Artsob, J. L. Pelegrino, P. Buchy, M. J. Cardosa, S. Devi, D.
A. Enria, J. Farrar, D. J. Gubler, M. G. Guzman, S. B. Halstead, E. Hunsperger,
S. Kliks, H. S. Margolis, C. M. Nathanson, V. C. Nguyen, N. Rizzo, S.
Vázquez, and S. Yoksan, “Evaluation of diagnostic tests: dengue,” Nat. Rev.
Microbiol., vol. 8, no. 12, pp. S30–S37, Dec. 2010.
[32] A. Sharwani, W. Jerjes, V. Salih, B. Swinson, I. J. Bigio, M. El-Maaytah, and
C. Hopper, “Assessment of oral premalignancy using elastic scattering
spectroscopy,” Oral Oncol., vol. 42, no. 4, pp. 343–349, Apr. 2006.
[33] W. K. Jerjes, T. Upile, B. J. Wong, C. S. Betz, H. J. Sterenborg, M. J. Witjes,
K. Berg, R. van Veen, M. A. Biel, A. K. El-Naggar, C. A. Mosse, M. Olivo, R.
Richards-Kortum, D. J. Robinson, J. Rosen, A. G. Yodh, C. Kendall, J. F.
Ilgner, A. Amelink, V. Bagnato, H. Barr, L. Bolotine, I. Bigio, Z. Chen, L.-P.
Choo-Smith, A. K. D’Cruz, A. Gillenwater, A. Leunig, A. J. MacRobert, G.
70
McKenzie, A. Sandison, K. C. Soo, H. Stepp, N. Stone, K. Svanberg, I. B. Tan,
B. C. Wilson, H. Wolfsen, and C. Hopper, “The future of medical diagnostics:
review paper.,” Head Neck Oncol., vol. 3, p. 38, Aug. 2011.
[34] W. Jerjes, B. Swinson, D. Pickard, G. . Thomas, and C. Hopper, “Detection of
cervical intranodal metastasis in oral cancer using elastic scattering
spectroscopy,” Oral Oncol., vol. 40, no. 7, pp. 673–678, Aug. 2004.
[35] R. Richards-Kortum and E. Sevick-Muraca, “QUANTITATIVE OPTICAL
SPECTROSCOPY FOR TISSUE DIAGNOSIS,” Annu. Rev. Phys. Chem., vol.
47, no. 1, pp. 555–606, Oct. 1996.
[36] S. He, K. L. Dunn, P. S. Espino, B. Drobic, L. Li, J. Yu, J.-M. Sun, H. Y. Chen,
S. Pritchard, and J. R. Davie, “Chromatin organization and nuclear
microenvironments in cancer cells,” J. Cell. Biochem., vol. 104, no. 6, pp.
2004–2015, Aug. 2008.
[37] A. Amelink, H. J. C. M. Sterenborg, M. P. L. Bard, and S. A. Burgers, “In vivo
measurement of the local optical properties of tissue by use of differential path-
length spectroscopy.,” Opt. Lett., vol. 29, no. 10, pp. 1087–9, May 2004.
[38] R. L. P. van Veen, A. Amelink, M. Menke-Pluymers, C. van der Pol, and H. J.
C. M. Sterenborg, “Optical biopsy of breast tissue using differential path-length
spectroscopy,” Phys. Med. Biol., vol. 50, no. 11, pp. 2573–2581, Jun. 2005.
[39] M. P. L. Bard, A. Amelink, M. Skurichina, V. Noordhoek Hegt, R. P. W. Duin,
H. J. C. M. Sterenborg, H. C. Hoogsteden, and J. G. J. V. Aerts, “Optical
Spectroscopy for the Classification of Malignant Lesions of the Bronchial
Tree,” Chest, vol. 129, no. 4, pp. 995–1001, Apr. 2006.
[40] A. Amelink, O. P. Kaspers, H. J. C. M. Sterenborg, J. E. van der Wal, J. L. N.
Roodenburg, and M. J. H. Witjes, “Non-invasive measurement of the
morphology and physiology of oral mucosa by use of optical spectroscopy,”
Oral Oncol., vol. 44, no. 1, pp. 65–71, Jan. 2008.
[41] T. Upile, W. Jerjes, H. J. Sterenborg, A. K. El-Naggar, A. Sandison, M. J.
Witjes, M. A. Biel, I. Bigio, B. J. Wong, A. Gillenwater, A. J. MacRobert, D. J.
Robinson, C. S. Betz, H. Stepp, L. Bolotine, G. McKenzie, C. Mosse, H. Barr,
Z. Chen, K. Berg, A. K. D’Cruz, N. Stone, C. Kendall, S. Fisher, A. Leunig, M.
Olivo, R. Richards-Kortum, K. Soo, V. Bagnato, L.-P. Choo-Smith, K.
Svanberg, I. B. Tan, B. C. Wilson, H. Wolfsen, A. G. Yodh, and C. Hopper,
“Head & neck optical diagnostics: vision of the future of surgery,” Head
Neck Oncol., vol. 1, no. 1, p. 25, Jul. 2009.
[42] L. Shan, Y. Hao, S. Wang, A. Korotcov, R. Zhang, T. Wang, J. Califano, X.
Gu, R. Sridhar, Z. M. Bhujwalla, and P. C. Wang, “Visualizing head and neck
tumors in vivo using near-infrared fluorescent transferrin conjugate.,” Mol.
Imaging, vol. 7, no. 1, pp. 42–9.
[43] T. Upile, W. Jerjes, C. S. Betz, M. El Maaytah, A. Wright, and C. Hopper,
“Optical diagnostic techniques in the head and neck.,” Dent. Update, vol. 34,
no. 7, p. 410–2, 415–6, 419–20 passim, Sep. 2007.
71
[44] R. Ullah, S. Khan, M. Bilal, F. Nurjis, and M. Saleem, “Non-invasive
assessment of mango ripening using fluorescence spectroscopy,” Optik
(Stuttg)., vol. 127, no. 13, pp. 5186–5189, 2016.
[45] R. Ullah, S. Khan, H. Ali, M. Bilal, and M. Saleem, “Identification of cow and
buffalo milk based on Beta carotene and vitamin-A concentration using
fluorescence spectroscopy,” PLoS One, vol. 12, no. 5, pp. 1–8, 2017.
[46] C. Arens, D. Reußner, J. Woenkhaus, A. Leunig, C. S. Betz, and H. Glanz,
“Indirect fluorescence laryngoscopy in the diagnosis of precancerous and
cancerous laryngeal lesions,” Eur. Arch. Oto-Rhino-Laryngology, vol. 264, no.
6, pp. 621–626, May 2007.
[47] A. Weber, Raman Spectroscopy of Gases and Liquids. Springer Berlin
Heidelberg, 1979.
[48] D. A. Long, The Raman effect : a unified treatment of the theory of Raman
scattering by molecules. Wiley, 2002.
[49] N. B. Colthup, L. H. Daly, and S. E. Wiberley, Introduction to infrared and
Raman spectroscopy. Academic Press, 1990.
[50] M. Bilal, M. Saleem, M. Bilal, M. Khurram, S. Khan, R. Ullah, H. Ali, and M.
Ahmed, “Raman spectroscopy based discrimination of NS1 positive and
negative dengue virus infected serum,” Laser Phys. Lett., vol. 13, no. 9, p.
95603, Sep. 2016.
[51] M. Bilal, M. Saleem, M. Bilal, T. Ijaz, S. Khan, R. Ullah, A. Raza, M.
Khurram, W. Akram, and M. Ahmed, “Raman spectroscopy-based screening of
IgM positive and negative sera for dengue virus infection,” Laser Phys., vol.
26, no. 11, p. 115602, 2016.
[52] M. Bilal, M. Saleem, M. Bial, S. Khan, R. Ullah, H. Ali, M. Ahmed, and M.
Ikram, “Raman spectroscopy based screening of IgG positive and negative sera
for dengue virus infection,” Laser Phys. Lett., vol. 14, no. 11, p. 115601, Nov.
2017.
[53] M. Bilal, R. Ullah, S. Khan, H. Ali, M. Saleem, and M. Ahmed, “Lactate based
optical screening of dengue virus infection in human sera using Raman
spectroscopy,” Biomed. Opt. Express, vol. 8, no. 2, p. 1250, Feb. 2017.
[54] D. J. Gubler, “The global emergence/resurgence of arboviral diseases as public
health problems,” Arch. Med. Res., vol. 33, no. 4, pp. 330–342, 2002.
[55] O. Parkash and R. H. Shueb, “Diagnosis of Dengue Infection Using
Conventional and Biosensor Based Techniques.,” Viruses, vol. 7, no. 10, pp.
5410–27, Oct. 2015.
[56] P.-Y. Shu and J.-H. Huang, “Current Advances in Dengue Diagnosis,” Clin.
Vaccine Immunol., vol. 11, no. 4, pp. 642–650, Jul. 2004.
[57] W. Dejnirattisai, A. Jumnainsong, N. Onsirisakul, P. Fitton, S. Vasanawathana,
W. Limpitikul, C. Puttikhunt, C. Edwards, T. Duangchinda, S. Supasa, K.
Chawansuntati, P. Malasit, J. Mongkolsapaya, and G. Screaton, “Cross-
72
Reacting Antibodies Enhance Dengue Virus Infection in Humans,” Science
(80-. )., vol. 328, no. 5979, pp. 745–748, May 2010.
[58] S. C. Kliks, A. Nisalak, W. E. Brandt, L. Wahl, and D. S. Burke, “Antibody-
dependent enhancement of dengue virus growth in human monocytes as a risk
factor for dengue hemorrhagic fever.,” Am. J. Trop. Med. Hyg., vol. 40, no. 4,
pp. 444–51, Apr. 1989.
[59] D. M. Morens and S. B. Halstead, “Measurement of Antibody-dependent
Infection Enhancement of Four Dengue Virus Serotypes by Monoclonal and
Polyclonal Antibodies,” J. Gen. Virol., vol. 71, no. 12, pp. 2909–2914, Dec.
1990.
[60] A. Sabchareon, J. Lang, P. Chanthavanich, S. Yoksan, R. Forrat, P. Attanath,
C. Sirivichayakul, K. Pengsaa, C. Pojjaroen-Anant, L. Chambonneau, J.
Saluzzo, A. Sabchareon, J. Lang, P. Chanthavanich, S. Yoksan, R. Forrat, P.
Attanath, C. Sirivichayakul, K. Pengsaa, C. Pojjaroen-Anant, L. Chambonneau,
J. F. Saluzzo, and N. Bhamarapravati, “Safety and immunogenicity of a three
dose regimen of two tetravalent live-attenuated dengue vaccines in five- to
twelve-year-old Thai children,” Pediatr. Infect. Dis. J., vol. 23, no. 2, pp. 99–
109, Feb. 2004.
[61] D. I. Mackenzie, L. L. Bailey, and J. D. Nichols, “Investigating species co-
occurrence patterns when species,” J. Anim. Ecol., vol. 73, no. 73, pp. 546–
555, 2004.
[62] C. Pancharoen, J. Mekmullica, and U. Thisyakorn, “Primary dengue infection:
What are the clinical distinctions from secondary infection?,” Southeast Asian
J. Trop. Med. Public Health, vol. 32, no. 3, pp. 476–480, Sep. 2001.
[63] S. W. Gollins and J. S. Porterfield, “Flavivirus infection enhancement in
macrophages: Radioactive and biological studies on the effect of antibody on
viral fate,” J. Gen. Virol., vol. 65, no. 8, pp. 1261–1272, Aug. 1984.
[64] A. Igarashi, “Isolation of a Singh’s Aedes albopictus Cell Clone Sensitive to
Dengue and Chikungunya Viruses,” J. Gen. Virol., vol. 40, no. 3, pp. 531–544,
Sep. 1978.
[65] M. Parida, K. Horioke, H. Ishida, P. K. Dash, P. Saxena, A. M. Jana, M. A.
Islam, S. Inoue, N. Hosaka, and K. Morita, “Rapid Detection and
Differentiation of Dengue Virus Serotypes by a Real-Time Reverse
Transcription-Loop-Mediated Isothermal Amplification Assay,” J. Clin.
Microbiol., vol. 43, no. 6, pp. 2895–2903, Jun. 2005.
[66] R. S. Lanciotti, C. H. Calisher, D. J. Gubler, G. J. Chang, and A. V. Vorndam,
“Rapid detection and typing of dengue viruses from clinical samples by using
reverse transcriptase-polymerase chain reaction,” J. Clin. Microbiol., vol. 30,
no. 3, pp. 545–551, Mar. 1992.
[67] C. Klungthong, R. V. Gibbons, B. Thaisomboonsuk, A. Nisalak, S.
Kalayanarooj, V. Thirawuth, N. Nutkumhang, M. P. Mammen, and R. G.
Jarman, “Dengue virus detection using whole blood for reverse transcriptase
PCR and virus isolation,” J. Clin. Microbiol., vol. 45, no. 8, pp. 2480–2485,
73
Aug. 2007.
[68] I. Yamada, K., Takasaki, T., Nawa, M. and Kurane, “Laboratory diagnosis of
imported dengue cases,” Jpn. J. Trop. Med. Hyg., vol. 27, pp. 75–77, 1999.
[69] K. Yamada, M. Nawa, T. Takasaki, S. Yabe, and I. Kurane, “Laboratory
diagnosis of dengue virus infection by reverse transcriptase polymerase chain
reaction (RT-PCR) and IgM-capture enzyme-linked immunosorbent assay
(ELISA),” Jpn J Infect Dis, vol. 52, no. 4, pp. 150–155, 1999.
[70] K. Bundo and A. Igarashi, “Antibody-capture ELISA for detection of
immunoglobulin M antibodies in sera from Japanese encephalitis and dengue
hemorrhagic fever patients.,” J. Virol. Methods, vol. 11, no. 1, pp. 15–22, May
1985.
[71] D. H. CLARKE and J. CASALS, “Techniques for hemagglutination and
hemagglutination-inhibition with arthropod-borne viruses.,” Am. J. Trop. Med.
Hyg., vol. 7, no. 5, pp. 561–73, Sep. 1958.
[72] E. A. Hunsperger, S. Yoksan, P. Buchy, V. C. Nguyen, S. D. Sekaran, D. A.
Enria, J. L. Pelegrino, S. Vázquez, H. Artsob, M. Drebot, D. J. Gubler, S. B.
Halstead, M. G. Guzmán, H. S. Margolis, C. M. Nathanson, N. R. R. Lic, K. E.
Bessoff, S. Kliks, and R. W. Peeling, “Evaluation of commercially available
anti-dengue virus immunoglobulin M tests,” Emerg. Infect. Dis., vol. 15, no. 3,
pp. 436–440, Mar. 2009.
[73] C. B. Berde, G. A. Walco, E. J. Krane, K. J. S. Anand, J. V. Aranda, K. D.
Craig, C. D. Dampier, J. C. Finkel, M. Grabois, C. Johnston, J. Lantos, A.
Lebel, L. G. Maxwell, P. McGrath, T. F. Oberlander, L. E. Schanberg, B.
Stevens, A. Taddio, C. L. von Baeyer, M. Yaster, and W. T. Zempsky,
“Pediatric Analgesic Clinical Trial Designs, Measures, and Extrapolation:
Report of an FDA Scientific Workshop,” Pediatrics, vol. 129, no. 2, pp. 354–
364, Feb. 2012.
[74] S. D. Blacksell, R. G. Jarman, R. V. Gibbons, A. Tanganuchitcharnchai, M. P.
Mammen, A. Nisalak, S. Kalayanarooj, M. S. Bailey, R. Premaratna, H. J. De
Silva, N. P. J. Day, and D. G. Lalloog, “Comparison of seven commercial
antigen and antibody enzyme-linked immunosorbent assays for detection of
acute dengue infection,” Clin. Vaccine Immunol., vol. 19, no. 5, pp. 804–810,
May 2012.
[75] A. Vazquez, A. Flammini, A. Maritan, and A. Vespignani, “Global protein
function prediction from protein-protein interaction networks,” Nat.
Biotechnol., vol. 21, no. 6, pp. 697–700, Jun. 2003.
[76] A. WHO Regional Office for South-East, Comprehensive Guidelines for
Prevention and Control of Dengue and Dengue Haemorrhagic Fever. WHO,
2011.
[77] World Health Organization (WHO) and the Special Programme for Research
and Tropical Diseases (TDR), Dengue Guidelines for diagnosis, tretment,
prevention and control. World Health Organization, 2009.
74
[78] C. Åström, J. Rocklöv, S. Hales, A. Béguin, V. Louis, and R. Sauerborn,
“Potential Distribution of Dengue Fever Under Scenarios of Climate Change
and Economic Development,” Ecohealth, vol. 9, no. 4, pp. 448–454, Dec.
2012.
[79] S. Hales, N. de Wet, J. Maindonald, and A. Woodward, “Potential effect of
population and climate changes on global distribution of dengue fever: an
empirical model,” Lancet, vol. 360, no. 9336, pp. 830–834, Sep. 2002.
[80] A. Wilder-Smith, E.-E. Ooi, S. G. Vasudevan, and D. J. Gubler, “Update on
Dengue: Epidemiology, Virus Evolution, Antiviral Drugs, and Vaccine
Development,” Curr. Infect. Dis. Rep., vol. 12, no. 3, pp. 157–164, May 2010.
[81] N. E. A. Murray, M. B. Quam, and A. Wilder-Smith, “Epidemiology of
dengue: past, present and future prospects.,” Clin. Epidemiol., vol. 5, pp. 299–
309, 2013.
[82] J. P. Messina, O. J. Brady, T. W. Scott, C. Zou, D. M. Pigott, K. A. Duda, S.
Bhatt, L. Katzelnick, R. E. Howes, K. E. Battle, C. P. Simmons, and S. I. Hay,
“Global spread of dengue virus types: mapping the 70 year history.,” Trends
Microbiol., vol. 22, no. 3, pp. 138–46, Mar. 2014.
[83] A. Wilder-Smith and D. J. Gubler, “Geographic Expansion of Dengue: The
Impact of International Travel,” Med. Clin. North Am., vol. 92, no. 6, pp.
1377–1390, Nov. 2008.
[84] A. Smekal, “Zur Quantentheorie der Streuung und Dispersion,”
Naturwissenschaften, vol. 16, no. 31, pp. 612–613, Oct. 1928.
[85] G. Landsberg and L. Mandelstam, “Eine neue Erscheinung bei der
Lichtzerstreuung in Krystallen,” Die Naturwissenschaften, vol. 16, no. 28. pp.
557–558, 1928.
[86] C. V. RAMAN and K. S. KRISHNAN, “A New Type of Secondary
Radiation,” Nature, vol. 121, no. 3048, pp. 501–502, 1928.
[87] R. Aroca, “Theory of Molecular Vibrations. The Origin of Infrared and Raman
Spectra,” in Surface-Enhanced Vibrational Spectroscopy, Chichester, UK: John
Wiley & Sons, Ltd, 2007, pp. 1–33.
[88] E. C. Le Ru, E. Blackie, M. Meyer, and P. G. Etchegoint, “Surface enhanced
raman scattering enhancement factors: A comprehensive study,” J. Phys.
Chem. C, vol. 111, no. 37, pp. 13794–13803, 2007.
[89] F. Légaré, C. L. Evans, F. Ganikhanov, X. S. Xie, W. Seitz, D. Kopf, J. Segall,
J. Pollard, J. Condeelis, R. Weinberg, and S. Morton, “Towards CARS
Endoscopy,” Opt. Express, vol. 14, no. 10, p. 4427, Jan. 2006.
[90] † Songzhou Hu, ‡ and Kevin M. Smith, and † Thomas G. Spiro*, “Assignment
of Protoheme Resonance Raman Spectrum by Heme Labeling in Myoglobin,”
1996.
[91] M. Tanaka and R. J. Young, “Review Polarised Raman spectroscopy for the
study of molecular orientation distributions in polymers,” J. Mater. Sci., vol.
75
41, no. 3, pp. 963–991, Feb. 2006.
[92] Q. Tu and C. Chang, “Diagnostic applications of Raman spectroscopy,”
Nanomedicine Nanotechnology, Biol. Med., vol. 8, no. 5, pp. 545–558, 2012.
[93] I. Pence and A. Mahadevan-jansen, “Clinical instrumentation and applications
of Raman spectroscopy,” Chem. Soc. Rev., vol. 45, no. 7, pp. 1958–1979, 2016.
[94] R. L. Frost, “A Raman spectroscopic study of selected minerals of the rosasite
group,” J. Raman Spectrosc., vol. 37, no. 9, pp. 910–921, 2006.
[95] D. C. G. de Veld, T. C. Bakker Schut, M. Skurichina, M. J. H. Witjes, J. E. Van
der Wal, J. L. N. Roodenburg, and H. J. C. M. Sterenborg, “Autofluorescence
and Raman microspectroscopy of tissue sections of oral lesions,” Lasers Med.
Sci., vol. 19, no. 4, pp. 203–209, Apr. 2005.
[96] A. S. Haka, Z. Volynskaya, J. A. Gardecki, J. Nazemi, R. Shenk, N. Wang, R.
R. Dasari, M. Fitzmaurice, and M. S. Feld, “Diagnosing breast cancer using
Raman spectroscopy: prospective analysis,” J. Biomed. Opt., vol. 14, no. 5, p.
54023, 2009.
[97] K. Kalyan Kumar, M. V. P. Chowdary, S. Mathew, L. Rao, C. Murali Krishna,
and J. Kurien, “Raman spectroscopic diagnosis of breast cancers: evaluation of
models,” J. Raman Spectrosc., vol. 39, no. 9, pp. 1276–1282, Sep. 2008.
[98] M. V. P. Chowdary, K. K. Kumar, J. Kurien, S. Mathew, and C. M. Krishna,
“Discrimination of normal, benign, and malignant breast tissues by Raman
spectroscopy,” Biopolymers, vol. 83, no. 5, pp. 556–569, Dec. 2006.
[99] M. Bilal, M. Bilal, S. Tabassum, M. Saleem, H. Mahmood, U. Sarwar, H.
Bangush, F. Munir, M. A. Zia, M. Ahmed, S. Shahzada, and E. U. Khan,
“Optical Screening of Female Breast Cancer from Whole Blood Using Raman
Spectroscopy,” Appl. Spectrosc., vol. 71, no. 5, pp. 1004–1013, 2016.
[100] A. T. Harris, A. Lungari, C. J. Needham, S. L. Smith, M. A. Lones, S. E.
Fisher, X. B. Yang, N. Cooper, J. Kirkham, D. A. Smith, D. P. Martin-Hirsch,
and A. S. High, “Potential for Raman spectroscopy to provide cancer screening
using a peripheral blood sample,” Head Neck Oncol., vol. 1, no. 1, p. 34, 2009.
[101] G. Shetty, C. Kendall, N. Shepherd, N. Stone, and H. Barr, “Raman
spectroscopy: elucidation of biochemical changes in carcinogenesis of
oesophagus.,” Br. J. Cancer, vol. 94, no. 10, pp. 1460–4, May 2006.
[102] A. T. Harris, M. Garg, X. B. Yang, S. E. Fisher, J. Kirkham, D. A. Smith, D. P.
Martin-Hirsch, and A. S. High, “Raman spectroscopy and advanced
mathematical modelling in the discrimination of human thyroid cell lines.,”
Head Neck Oncol., vol. 1, p. 38, Oct. 2009.
[103] S. Khan, R. Ullah, S. Javaid, S. Shahzad, H. Ali, M. Bilal, M. Saleem, and M.
Ahmed, “Raman Spectroscopy Combined with Principal Component Analysis
for Screening Nasopharyngeal Cancer in Human Blood Sera,” Appl. Spectrosc.,
p. 370281772392, Aug. 2017.
[104] P. C. Ashok, G. P. Singh, K. M. Tan, and K. Dholakia, “Fiber probe based
76
microfluidic raman spectroscopy.,” Opt. Express, vol. 18, no. 8, pp. 7642–9,
Apr. 2010.
[105] S. Dochow, M. Becker, R. Spittel, C. Beleites, S. Stanca, I. Latka, K. Schuster,
J. Kobelke, S. Unger, T. Henkel, G. Mayer, J. Albert, M. Rothhardt, C. Krafft,
and J. Popp, “Raman-on-chip device and detection fibres with fibre Bragg
grating for analysis of solutions and particles.,” Lab Chip, vol. 13, no. 6, pp.
1109–13, Mar. 2013.
[106] M. Sánchez-Purrà, M. Carré-Camps, H. de Puig, I. Bosch, L. Gehrke, and K.
Hamad-Schifferli, “Surface-Enhanced Raman Spectroscopy-Based Sandwich
Immunoassays for Multiplexed Detection of Zika and Dengue Viral
Biomarkers,” ACS Infect. Dis., vol. 3, no. 10, pp. 767–776, Oct. 2017.
[107] A. M. Paul, Z. Fan, S. S. Sinha, Y. Shi, L. Le, F. Bai, and P. C. Ray,
“Bioconjugated Gold Nanoparticle Based SERS Probe for Ultrasensitive
Identification of Mosquito-Borne Viruses Using Raman Fingerprinting,” J.
Phys. Chem. C, vol. 119, no. 41, pp. 23669–23675, Oct. 2015.
[108] R. AnandaRao, S. Swaminathan, S. Fernando, A. M. Jana, and N. Khanna, “A
custom-designed recombinant multiepitope protein as a dengue diagnostic
reagent,” Protein Expr. Purif., vol. 41, no. 1, pp. 136–147, May 2005.
[109] L. A. C. Burnett, G. Lunn, and R. Coico, “Biosafety: Guidelines for working
with pathogenic and infectious microorganisms,” Current Protocols in
Microbiology, no. SUPPL. 13. 2009.
[110] E. V. Thomas and D. M. Haaland, “Comparison of multivariate calibration
methods for quantitative spectral analysis,” Anal. Chem., vol. 62, no. 10, pp.
1091–1099, May 1990.
[111] L. Pasti, B. Walczak, D. L. Massart, and P. Reschiglian, “Optimization of
signal denoising in discrete wavelet transform,” Chemom. Intell. Lab. Syst.,
vol. 48, no. 1, pp. 21–34, 1999.
[112] P. Lasch, “Spectral pre-processing for biomedical vibrational spectroscopy and
microspectroscopic imaging,” Chemom. Intell. Lab. Syst., vol. 117, no. 1
August 2012, pp. 100–114, 2012.
[113] Q. Wang, “Raman spectroscopic characterization and analysis of agricultural
and biological systems,” Graduate Theses and Dissertations (Iowa State
University). p. 157, 2013.
[114] S. Utsunomiya, Y. Fujita, S. Tanaka, S. Kajihara, K. Aoshima, Y. Oda, and K.
Tanaka, “Signal Processing Algorithm Development for Mass++ (Ver. 2):
Platform Software for Mass Spectrometry,” IPSJ Trans. Bioinforma., vol. 7,
pp. 24–29, 2014.
[115] R. Galli, O. Uckermann, E. Koch, G. Schackert, M. Kirsch, and G. Steiner,
“Effects of tissue fixation on coherent anti-Stokes Raman scattering images of
brain,” J. Biomed. Opt., vol. 19, no. 7, p. 71402, 2013.
[116] S. Wold, M. Sjöström, and L. Eriksson, “PLS-regression: A basic tool of
77
chemometrics,” Chemom. Intell. Lab. Syst., vol. 58, no. 2, pp. 109–130, 2001.
[117] O. Yeniay and A. Goktas, “A Comparison of Partial Least Squares Regression
with Other Prediction Methods,” J. Math. Stat., vol. 31, pp. 99–111, 2002.
[118] R. G. Brereton, “Introduction to multivariate calibration in analytical
chemistry,” Analyst, vol. 125, no. 11, pp. 2125–2154, Jan. 2000.
[119] S. De Jong, “Multivariate calibration, H. Martens and T. Naes, Wiley, New
York, 1989. ISBN 0 471 90979 3. Price: £75.00, US$138.00. No. of pages:
504,” J. Chemom., vol. 4, no. 6, pp. 441–441, Nov. 1990.
[120] R. Bro and A. K. Smilde, “Principal component analysis,” Anal. Methods, vol.
6, no. 9, pp. 2812–2831, 2014.
[121] T. Næs and H. Martens, “Multivariate calibration. II. Chemometric methods,”
Trends Anal. Chem., vol. 3, no. 10, pp. 266–271, 1984.
[122] B. Cochran and C. Logue, “A Watershed Approach to Improve Water Quality:
Case Study of Clean Water Services’ Tualatin River Program,” J. Am. Water
Resour. Assoc., vol. 47, no. 1, pp. 29–38, 2011.
[123] K. Hajian-Tilaki, “Receiver operating characteristic (ROC) curve analysis for
medical diagnostic test evaluation,” Casp. J. Intern. Med., vol. 4, no. 2, pp.
627–635, 2013.
[124] V. Nykänen, I. Lahti, T. Niiranen, and K. Korhonen, “Receiver operating
characteristics (ROC) as validation tool for prospectivity models - A magmatic
Ni-Cu case study from the Central Lapland Greenstone Belt, Northern
Finland,” Ore Geol. Rev., vol. 71, no. December 2015, pp. 853–860, 2015.
[125] M. H. Zweig and G. Campbell, “Receiver-operating characteristic (ROC) plots:
A fundamental evaluation tool in clinical medicine,” Clin. Chem., vol. 39, no.
4, pp. 561–577, 1993.
[126] J. De Gelder, K. De Gussem, P. Vandenabeele, and L. Moens, “Reference
database of Raman spectra of biological molecules,” J. Raman Spectrosc., vol.
38, no. 9, pp. 1133–1147, 2007.
[127] A. C. S. Talari, Z. Movasaghi, S. Rehman, and I. ur Rehman, “Raman
Spectroscopy of Biological Tissues,” Appl. Spectrosc. Rev., vol. 50, no. 1, pp.
46–111, 2015.
[128] C. Krafft, L. Neudert, T. Simat, and R. Salzer, “Near infrared Raman spectra of
human brain lipids.,” Spectrochim. Acta. A. Mol. Biomol. Spectrosc., vol. 61,
no. 7, pp. 1529–35, May 2005.
[129] R. Perera, C. Riley, G. Isaac, A. S. Hopf-Jannasch, R. J. Moore, K. W. Weitz,
L. Pasa-Tolic, T. O. Metz, J. Adamec, and R. J. Kuhn, “Dengue virus infection
perturbs lipid homeostasis in infected mosquito cells,” PLoS Pathog., vol. 8,
no. 3, p. e1002584, Mar. 2012.
[130] M. G. Jacobs, P. J. Robinson, C. Bletchly, J. M. Mackenzie, and P. R. Young,
“Dengue virus nonstructural protein 1 is expressed in a glycosyl-
78
phosphatidylinositol-linked form that is capable of signal transduction,” Faseb
J, vol. 14, no. 11, pp. 1603–1610, Aug. 2000.
[131] H. Wang, N. Huang, J. Zhao, H. Lui, M. Korbelik, and H. Zeng, “Depth-
resolved in vivo micro-Raman spectroscopy of a murine skin tumor model
reveals cancer-specific spectral biomarkers,” J. Raman Spectrosc., vol. 42, no.
2, pp. 160–166, Feb. 2011.
[132] L. Cui, Y. H. Lee, Y. Kumar, F. Xu, K. Lu, E. E. Ooi, S. R. Tannenbaum, and
C. N. Ong, “Serum metabolome and lipidome changes in adult patients with
primary dengue infection.,” PLoS Negl. Trop. Dis., vol. 7, no. 8, p. e2373, Jan.
2013.
[133] S. Farquharson, C. Shende, F. E. Inscore, P. Maksymiuk, and A. Gift,
“Analysis of 5-fluorouracil in saliva using surface-enhanced Raman
spectroscopy,” J. Raman Spectrosc., vol. 36, no. 3, pp. 208–212, Mar. 2005.
[134] J. D. Chandler and B. J. Day, “Thiocyanate: A potentially useful therapeutic
agent with host defense and antioxidant properties,” Biochem. Pharmacol., vol.
84, no. 11, pp. 1381–1387, 2012.
[135] L. Silveira, S. Sathaiah, R. A. Zngaro, M. T. T. Pacheco, M. C. Chavantes, and
C. A. G. Pasqualucci, “Correlation between near-infrared Raman spectroscopy
and the histopathological analysis of atherosclerosis in human coronary
arteries,” Lasers Surg. Med., vol. 30, no. 4, pp. 290–297, 2002.
[136] S. Kaminaka, T. Ito, H. Yamazaki, E. Kohda, and H. Hamaguchi, “Near-
infrared multichannel Raman spectroscopy toward real-timein vivo cancer
diagnosis,” J. Raman Spectrosc., vol. 33, no. 7, pp. 498–502, Jul. 2002.
[137] A. J. D. Moreno, P. T. C. Freire, I. Guedes, F. E. A. Melo, J. Mendes-Filho,
and J. A. Sanjurjo, “Raman scattering studies of monohydrated L-asparagine,”
Brazilian J. Phys., vol. 29, no. 2, pp. 380–387, Jun. 1999.
[138] J. A. Mondotte, P.-Y. Lozach, A. Amara, and A. V. Gamarnik, “Essential Role
of Dengue Virus Envelope Protein N Glycosylation at Asparagine-67 during
Viral Propagation,” J. Virol., vol. 81, no. 13, pp. 7136–7148, Jul. 2007.
[139] F. Chiappelli, S. M. E. Santos, X. M. Caldeira Brant, A. Bakhordarian, A. D.
Thames, C. A. Maida, A. M. Du, A. L. Jan, M. Nahcivan, M. T. Nguyen, and
N. Sama, “Viral immune evasion in dengue: toward evidence-based revisions
of clinical practice guidelines.,” Bioinformation, vol. 10, no. 12, pp. 726–33,
Jan. 2014.
[140] F. Melchers, “Difference in Carbohydrate Composition and a Possible
Conformational Difference between Intracellular and Extracellular
Immunoglobulin M,” Biochemistry, vol. 11, no. 11, pp. 2204–2208, May 1972.
[141] M. Maes, I. Mihaylova, M. Kubera, J. C. Leunis, and M. Geffard, “IgM-
mediated autoimmune responses directed against multiple neoepitopes in
depression: New pathways that underpin the inflammatory and
neuroprogressive pathophysiology,” J. Affect. Disord., vol. 135, no. 1–3, pp.
414–418, Dec. 2011.
79
[142] M. A. de Pablo and G. Alvarez de Cienfuegos, “Modulatory effects of dietary
lipids on immune system functions.,” Immunol. Cell Biol., vol. 78, no. 1, pp.
31–9, Feb. 2000.
[143] M. A. Hasanat, M. A. Ananna, M. U. Ahmed, and M. N. Alam, “Testing blood
glucose may be useful in the management of dengue.,” Mymensingh Med. J.,
vol. 19, no. 3, pp. 382–5, Jul. 2010.
[144] A. M. Turpeinen, M. Mutanen, A. Aro, I. Salminen, S. Basu, D. L. Palmquist,
and J. M. Griinari, “Bioconversion of vaccenic acid to conjugated linoleic acid
in humans.,” Am. J. Clin. Nutr., vol. 76, no. 3, pp. 504–10, Sep. 2002.
[145] K. S. Burrack and T. E. Morrison, “The role of myeloid cell activation and
arginine metabolism in the pathogenesis of virus-induced diseases.,” Front.
Immunol., vol. 5, p. 428, Jan. 2014.
[146] K. Song, H. Du, Q. Zhang, C. Wang, Y. Guo, H. Wu, L. Liu, Q. Jia, X. Wang,
H. Shi, S. Sun, and K. Niu, “Serum immunoglobulin M concentration is
positively related to metabolic syndrome in an adult population: Tianjin
Chronic Low-Grade Systemic Inflammation and Health (TCLSIH) Cohort
Study.,” PLoS One, vol. 9, no. 2, p. e88701, Jan. 2014.
[147] K. A. Fontaine, E. L. Sanchez, R. Camarda, and M. Lagunoff, “Dengue Virus
Induces and Requires Glycolysis for Optimal Replication,” J. Virol., vol. 89,
no. 4, pp. 2358–2366, Feb. 2015.
[148] X. Carnec, L. Meertens, O. Dejarnac, M. Perera-Lecoin, M. L. Hafirassou, J.
Kitaura, R. Ramdasi, O. Schwartz, and A. Amara, “The Phosphatidylserine and
Phosphatidylethanolamine Receptor CD300a Binds Dengue Virus and
Enhances Infection.,” J. Virol., vol. 90, no. 1, pp. 92–102, Jan. 2015.
[149] D. M. Chisenhall, R. C. Christofferson, M. K. McCracken, A.-M. F. Johnson,
B. Londono-Renteria, and C. N. Mores, “Infection with dengue-2 virus alters
proteins in naturally expectorated saliva of Aedes aegypti mosquitoes,”
Parasit. Vectors, vol. 7, no. 1, p. 252, Jan. 2014.
[150] A. Gonzalez-Quintela, R. Alende, F. Gude, J. Campos, J. Rey, L. M. Meijide,
C. Fernandez-Merino, and C. Vidal, “Serum levels of immunoglobulins (IgG,
IgA, IgM) in a general adult population and their relationship with alcohol
consumption, smoking and common metabolic abnormalities.,” Clin. Exp.
Immunol., vol. 151, no. 1, pp. 42–50, Jan. 2008.
[151] J. B. Henry, “Clinical diagnosis and management by laboratory methods. -
NLM Catalog - NCBI,” Philadelphia : Saunders, c1996, 1996. [Online].
Available: http://www.ncbi.nlm.nih.gov/nlmcatalog/9513812#. [Accessed: 02-
Aug-2016].
[152] C. W. Anderson and S. A. Rounds, “Use of Continuous Monitors and
Autosamplers to Predict Unmeasured Water-Quality Constituents in Tributaries
of the Tualatin River, Oregon,” U.S. Geol. Surv. Sci. Investig. Rep., vol. 5008,
p. 76, 2010.
[153] H. J. Staines, L. Garcia-Fernandez, R. Pogothata, P. S. Wallace, W. G.
80
MacKay, and a. M. Van Loon, “Monitoring performance of nucleic acid-based
diagnostic measurement system users by EQA,” Accredit. Qual. Assur., vol.
14, no. 5, pp. 243–252, 2009.
[154] A. Khedr, M. Hegazy, A. Kamal, and M. A. Shehata, “Profiling of esterified
fatty acids as biomarkers in the blood of dengue fever patients using a
microliter-scale extraction followed by gas chromatography and mass
spectrometry,” J. Sep. Sci., vol. 38, no. 2, pp. 316–324, Jan. 2015.
[155] W. Dejnirattisai, A. I. Webb, V. Chan, A. Jumnainsong, A. Davidson, J.
Mongkolsapaya, and G. Screaton, “Lectin switching during dengue virus
infection,” J. Infect. Dis., vol. 203, no. 12, pp. 1775–1783, Jun. 2011.
[156] W.-C. Tang, R.-J. Lin, C.-L. Liao, and Y.-L. Lin, “Rab18 facilitates dengue
virus infection by targeting fatty acid synthase to sites of viral replication.,” J.
Virol., vol. 88, no. 12, pp. 6793–804, 2014.
[157] V. Myriam, C. Maria, C. Sigrid, A. Sandra, P. Maryi, B. Edgar, and C. Jaime,
“In vivo and In vitro study of the Neuroinfection and Neuropathogeny induced
by dengue virus,” Front. Cell. Neurosci., vol. 9, 2015.
[158] J. M. (Jeremy M. Berg, J. L. Tymoczko, L. Stryer, and L. Stryer, Biochemistry,
5th ed. W.H. Freeman, 2002.
[159] S. Malhotra, N. Kaur, B. Ms, P. Kumar, and C. Hans, “Yeast Infection and
Psychiatric disorders,” Oct. 2010 DELHI PSYCHIATRY J. Psychomicrobiology
Delhi Psychiatry J., vol. 13, no. 2, pp. 345–350, 2010.
[160] S. G. Sankar, K. J. Dhanajeyan, R. Paramasivan, V. Thenmozhi, B. K. Tyagi,
and S. J. Vennison, “High-level expression of functionally active dengue-2
non-structural antigen 1 production in escherichia coli,” Biomed Res. Int., vol.
2013, p. 343195, 2013.
[161] E. E. Girard, A. Al-Ahmad, J. Rosenberg, R. Luong, T. Moore, G. Lauritsch, J.
Boese, and R. Fahrig, “Contrast-enhanced C-arm CT evaluation of
radiofrequency ablation lesions in the left ventricle,” JACC Cardiovasc.
Imaging, vol. 4, no. 3, pp. 259–268, 2011.
[162] A. Ströhle and A. Hahn, “[Vitamin C and immune function].,” Medizinische
Monatsschrift fur Pharm., vol. 32, no. 2, pp. 49-54–6, Feb. 2009.
[163] S. Goodman, Vitamin C the Master Nutrient, no. Library-Copy 1 Cuomo-Copy
2. New Canaan: Keats Publishing, 1991.
[164] S. A. Cross, D. R. Cook, A. W. S. Chi, P. J. Vance, L. L. Kolson, B. J. Wong,
K. L. Jordan-Sciutto, and D. L. Kolson, “Dimethyl Fumarate, an Immune
Modulator and Inducer of the Antioxidant Response, Suppresses HIV
Replication and Macrophage-Mediated Neurotoxicity: A Novel Candidate for
HIV Neuroprotection,” J. Immunol., vol. 187, no. 10, pp. 5015–5025, Nov.
2011.
[165] R. Perera, C. Riley, G. Isaac, A. S. Hopf-Jannasch, R. J. Moore, K. W. Weitz,
L. Pasa-Tolic, T. O. Metz, J. Adamec, and R. J. Kuhn, “Dengue virus infection
81
perturbs lipid homeostasis in infected mosquito cells,” PLoS Pathog., vol. 8,
no. 3, p. e1002584, Mar. 2012.
[166] P. Klassen, P. Fürst, C. Schulz, M. Mazariegos, and N. W. Solomons, “Plasma
free amino acid concentrations in healthy Guatemalan adults and in patients
with classic dengue,” Am. J. Clin. Nutr., vol. 73, no. 3, pp. 647–652, Mar.
2001.
[167] C. H. C. Dejong, M. C. G. van de Poll, P. B. Soeters, R. Jalan, and S. W. M.
Olde Damink, “Aromatic amino acid metabolism during liver failure.,” J.
Nutr., vol. 137, no. 6 Suppl 1, p. 1579S–1585S; discussion 1597S–1598S, Jun.
2007.
[168] Z. Movasaghi, S. Rehman, and I. U. Rehman, “Raman Spectroscopy of
Biological Tissues,” Appl. Spectrosc. Rev., vol. 42, no. 5, pp. 493–541, Sep.
2007.
[169] US Department of Health and Human Services, “Biosafety in Microbiological
and Biomedical Laboratories,” Public Heal. Serv., vol. 5th Editio, no. April,
pp. 1–250, 1999.
[170] L. W. Andersen, J. Mackenhauer, J. C. Roberts, K. M. Berg, M. N. Cocchi, and
M. W. Donnino, “Etiology and therapeutic approach to elevated lactate levels,”
Mayo Clin. Proc., vol. 88, no. 10, pp. 1127–1140, Oct. 2013.
[171] a Consoli, N. Nurjhan, J. J. Reilly, D. M. Bier, and J. E. Gerich, “Contribution
of liver and skeletal muscle to alanine and lactate metabolism in humans.,” Am.
J. Physiol., vol. 259, no. 5 Pt 1, pp. E677-84, Nov. 1990.
[172] G. Van Hall, “Lactate kinetics in human tissues at rest and during exercise,”
Acta Physiol., vol. 199, no. 4, pp. 499–508, Aug. 2010.
[173] S. Pilotto, M. T. Pacheco, L. Silveira, a B. Villaverde, and R. a Zângaro,
“Analysis of near-infrared Raman spectroscopy as a new technique for a
transcutaneous non-invasive diagnosis of blood components.,” Lasers Med.
Sci., vol. 16, no. 1, pp. 2–9, 2001.
[174] S. Kanai and T. Honda, “Liver function tests in patients with bacteremia,” J.
Clin. …, vol. 69, no. August 2007, pp. 66–69, 2008.
[175] T. L. Nguyen, T. H. Nguyen, and N. T. Tieu, “The impact of dengue
haemorrhagic fever on liver function,” Res. Virol., vol. 148, no. 4, pp. 273–277,
1997.
[176] S. L. Seneviratne, G. N. Malavige, and H. J. de Silva, “Pathogenesis of liver
involvement during dengue viral infections,” Trans. R. Soc. Trop. Med. Hyg.,
vol. 100, no. 7, pp. 608–614, Jul. 2006.
[177] A. Kabir, A. Abdullah, M. Sadeka, H. Ahmed, and M. Kahhar, “The Impact of
Dengue on Liver Function as Evaluated by Aminotransferase Levels,” J. Med.,
vol. 9, no. 2, pp. 66–68, Oct. 2008.
[178] T. F. Póvoa, A. M. B. Alves, C. A. B. Oliveira, G. J. Nuovo, V. L. A. Chagas,
and M. V. Paes, “The pathology of severe dengue in multiple organs of human
82
fatal cases: Histopathology, ultrastructure and virus replication,” PLoS One,
vol. 9, no. 4, 2014.
[179] P. Sirikutt and S. Kalayanarooj, “Serum lactate and lactate dehydrogenase as
parameters for the prediction of dengue severity,” J. Med. Assoc. Thail., vol.
97, pp. S220–S231, Jun. 2014.
[180] S. Gulati and A. Maheshwari, “Dengue fever-like illnesses: How different are
they from each other?,” Scand. J. Infect. Dis., vol. 44, no. 7, pp. 522–530, Jul.
2012.
top related