blind testing and evaluation of a comprehensive dna ...prediction results methods introduction blind...

1
Prediction Results Methods Introduction Blind Testing and Evaluation of a Comprehensive DNA Phenotyping System Rachel Wiley 1 , Xiangpei Zeng 1 , Bobby Larue 1 , Ellen M. Greytak 2 , Steven Armentrout 2 , Bruce Budowle 1,3 1) Institute of Applied Genetics, Department of Molecular and Medical Genetics, University of North Texas Health Science Center (UNTHSC), Forth Worth, TX; 2) Parabon NanoLabs, Inc., Reston, VA; 3) Center of Excellence in Genomic Medicine Research (CEGMR), King Abdulaziz University, Jeddah, Saudi Arabia Predictions Vs. Actual Appearance DNA phenotyping refers to the prediction of ancestry and/or physical appearance from DNA. In forensics, these predictions have the potential to generate new investigative leads in cases where DNA does not match a known suspect or a database, and to discover more information about unidentified remains. In this study, the Parabon ® Snapshot DNA Phenotyping System, which predicts detailed biogeographic ancestry, pigmentation (eye color, hair color, skin color, and freckling), and face morphology, was evaluated in a blind experiment. This study represents the first public blind evaluation of a comprehensive DNA phenotyping system, including side-by-side comparisons of the composite images and the actual photographs of each subject. 0% 50% 100% Blue Green Hazel Brown Black 0% 50% 100% Blond Brown Black Skin Color VFair Fair LtBrn Brown DkBrn Eye Color Green Brown Hair Color Blond Brown Black -Red +Red Freckles Zero Few Some Many Composite Actual Width Height Depth Regional Ancestry This study demonstrated the predictive performance of the Parabon Snapshot DNA Phenotyping system. Overall, the predicted features were consistent with the actual phenotypes: skin color, eye color, hair color, freckling, and ancestry. This phase of the study serves as a preliminary assessment of Level 1 detail so that strengths and limitations could be identified to set up a more in-depth analysis of face morphology in phase 2. Conclusions Brown* 0% 20% 40% 60% 80% 100% 23 - AfAm 24 - AfAm 16 - Jamaica 8 - India 21 - India 22 - N India 19 - S India 2 - China 12 - China 3 - Japan 18 - S China 17 - S China 14 - Europe 1 - Europe 7 - Europe 9 - Europe 10 - Europe 15 - Europe 20 - Europe 25 - Europe 5 - Europe 6 - Europe 4 - Europe 13 - Lebanon Predicted Ancestry Subject ID – Self-Reported Ancestry MidEast-W MidEast-N Europe-W Europe-SW Europe-SE Europe-S Europe-NE Europe-N EAsia-SE EAsia-Japan EAsia-Cen CAsia-S CAsia-NE CAsia-N CAsia-E Africa-W Green* Brown* Blond* ° § § 0% 50% 100% 1 2 3 4 5 6 7 8 9 10 12 13 14 15 16 17 18 19 20 21 22 23 24 25 VFair Fair LtBrn Brown DkBrn Fair* Fair* Predicted Phenotype Consistencies vs. Actual Phenotype 96% 96% 92% Skin Color Eye Color Hair Color Subject ID Number 24 subjects recruited for phenotypic and ancestral diversity by the University of North Texas Health Science Center (UNTHSC) 25 anonymous DNA samples sent to Parabon, including one two-person mixture (not made known to Parabon, but Parabon readily detected the mixture and identified the contributors) Each sample genotyped on the Illumina CytoSNP-850K chip (851,274 SNPs) and run through the Snapshot algorithms Phenotype predictions compiled into a detailed report for each subject, including a predicted composite in which differences from the average face for the same sex and ancestry were emphasized Age and body mass index (BMI) values then delivered to Parabon, and subjects with large differences from default age (25) and BMI (22) age-progressed by a forensic artist Photographs and self-reported ancestry and phenotypes collected by UNTHSC, and predictions for each Level 1 phenotype (sex, pigmentation, ancestry) compared to actual phenotypes Next phase will incorporate 3D scanning and craniofacial measurements to assess accuracy of predicted face morphology Study funded in part by the National Geographic Society Blue Hazel Black VFair Fair LtBrn Brown DkBrn Green Brown Blond Brown Black -Red +Red Zero Few Some Many Blue Hazel Black VFair Fair LtBrn DkBrn Green Blond Black -Red +Red Zero Few Some Many Blue Hazel Black Brown Brown Brown VFair Fair LtBrn DkBrn Green Blond Black -Red +Red Zero Few Some Many Blue Hazel Black Brown Brown Brown VFair Fair LtBrn DkBrn Green Blond Black -Red +Red Zero Few Some Many Blue Hazel Black Brown Brown Brown * Did not give permission to use image § Self-reported ° Blond as a young adult

Upload: others

Post on 29-Sep-2020

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Blind Testing and Evaluation of a Comprehensive DNA ...Prediction Results Methods Introduction Blind Testing and Evaluation of a Comprehensive DNA Phenotyping System Rachel Wiley1,

Prediction Results

Methods

Introduction

Blind Testing and Evaluation of a Comprehensive DNA Phenotyping System

Rachel Wiley1, Xiangpei Zeng1, Bobby Larue1, Ellen M. Greytak2, Steven Armentrout2, Bruce Budowle1,3

1) Institute of Applied Genetics, Department of Molecular and Medical Genetics, University of North Texas Health Science Center (UNTHSC), Forth Worth, TX;

2) Parabon NanoLabs, Inc., Reston, VA; 3) Center of Excellence in Genomic Medicine Research (CEGMR), King Abdulaziz University, Jeddah, Saudi Arabia

Predictions Vs. Actual Appearance DNA phenotyping refers to the prediction of ancestry and/or physical appearance from DNA. In forensics, these predictions have the potential to generate new investigative leads in cases

where DNA does not match a known suspect or a database, and to discover more information about unidentified remains. In this study,

the Parabon® Snapshot™ DNA Phenotyping System, which predicts detailed biogeographic ancestry, pigmentation (eye color,

hair color, skin color, and freckling), and face morphology, was evaluated in a blind experiment. This study represents the first public blind evaluation of a comprehensive DNA phenotyping

system, including side-by-side comparisons of the composite images and the actual photographs of each subject.

0% 50% 100%

Blue Green Hazel Brown Black

0% 50% 100%

Blond Brown Black

Skin Color

VF

air

Fair

LtB

rn

Bro

wn

DkB

rn

Eye Color

Gre

en

Bro

wn

Hair Color

Blo

nd

B

row

n

Bla

ck

-Red

+

Red

Freckles

Zero

F

ew

S

om

e

Many

Composite Actual

Width Height Depth Regional Ancestry

This study demonstrated the predictive performance of the Parabon Snapshot DNA Phenotyping system. Overall, the predicted features were consistent with the actual phenotypes:

skin color, eye color, hair color, freckling, and ancestry. This phase of the study serves as a preliminary assessment of Level 1 detail so

that strengths and limitations could be identified to set up a more in-depth analysis of face morphology in phase 2.

Conclusions

Brown*

0%

20%

40%

60%

80%

100%

23

- A

fAm

24

- A

fAm

16

- J

am

aic

a

8 -

Ind

ia

21

- Ind

ia

22

- N

Ind

ia

19

- S

Ind

ia

2 -

Chin

a

12

- C

hin

a

3 -

Jap

an

18

- S

Chin

a

17

- S

Chin

a

14

- E

uro

pe

1 -

Euro

pe

7 -

Euro

pe

9 -

Euro

pe

10

- E

uro

pe

15

- E

uro

pe

20

- E

uro

pe

25

- E

uro

pe

5 -

Euro

pe

6 -

Euro

pe

4 -

Euro

pe

13

- L

eb

anon

Pre

dic

ted

An

ce

str

y

Subject ID – Self-Reported Ancestry

MidEast-W MidEast-N Europe-W Europe-SW Europe-SE Europe-S Europe-NE Europe-N EAsia-SE EAsia-Japan EAsia-Cen CAsia-S CAsia-NE CAsia-N CAsia-E Africa-W

Green*

Brown*

Blond*

°

§

§

0% 50% 100%

1

2

3

4

5

6

7

8

9

10

12

13

14

15

16

17

18

19

20

21

22

23

24

25

VFair Fair LtBrn Brown DkBrn

Fair*

Fair*

Predicted Phenotype Consistencies vs. Actual Phenotype 96% 96% 92% Skin Color Eye Color Hair Color

Su

bje

ct

ID N

um

be

r

•  24 subjects recruited for phenotypic and ancestral diversity by the University of North Texas Health Science Center (UNTHSC)

•  25 anonymous DNA samples sent to Parabon, including one

two-person mixture (not made known to Parabon, but Parabon readily detected the mixture and identified the contributors)

•  Each sample genotyped on the Illumina CytoSNP-850K chip (851,274 SNPs) and run through the Snapshot algorithms

•  Phenotype predictions compiled into a detailed report for each subject, including a predicted composite in which differences from the average face for the same sex and ancestry were

emphasized •  Age and body mass index (BMI) values then delivered to

Parabon, and subjects with large differences from default age (25) and BMI (22) age-progressed by a forensic artist

•  Photographs and self-reported ancestry and phenotypes collected by UNTHSC, and predictions for each Level 1

phenotype (sex, pigmentation, ancestry) compared to actual phenotypes

•  Next phase will incorporate 3D scanning and craniofacial

measurements to assess accuracy of predicted face morphology

Study funded in part by the National Geographic Society

Blu

e

Haze

l

Bla

ck

VF

air

Fair

LtB

rn

Bro

wn

DkB

rn

Gre

en

Bro

wn

Blo

nd

B

row

n

Bla

ck

-Red

+

Red

Zero

F

ew

S

om

e

Many

Blu

e

Haze

l

Bla

ck

VFair

Fair

LtB

rn

DkB

rn

Gre

en

Blo

nd

Bla

ck

-Red

+

Red

Zero

F

ew

S

om

e

Many

Blu

e

Haze

l

Bla

ck

Bro

wn

Bro

wn

Bro

wn

VF

air

Fair

LtB

rn

DkB

rn

Gre

en

Blo

nd

Bla

ck

-Red

+

Red

Zero

F

ew

S

om

e

Many

Blu

e

Haze

l

Bla

ck

Bro

wn

Bro

wn

Bro

wn

VF

air

Fair

LtB

rn

DkB

rn

Gre

en

Blo

nd

Bla

ck

-Red

+

Red

Zero

Few

S

om

e

Many

Blu

e

Haze

l

Bla

ck

Bro

wn

Bro

wn

Bro

wn

* D

id n

ot

giv

e p

erm

issio

n t

o u

se im

ag

e

§ S

elf-r

ep

ort

ed

°

Blo

nd

as a

young

ad

ult