nsw forensic & analytical science service€¦ · • forensic biology & dna • forensic...
Post on 26-Apr-2020
32 Views
Preview:
TRANSCRIPT
21/09/2015
1
Two Years Later:
A reflection on the implementation of STRmix™ in a high throughput DNA laboratory
ISHI 2015
Zane Kerr
Senior Forensic Biologist
NSW Forensic & Analytical Science Service, Sydney, Australia
Outline
• Intro to NSW Forensic & Analytical Science Service
• Impetus for change
• Intro to STRmix™
• Validation
• Training
• Challenges
NSW Forensic & Analytical Science
Service
21/09/2015
2
A statewide forensic service provider
• Forensic Medicine
• Forensic Biology & DNA
• Forensic Toxicology
• Drug Toxicology
• Illicit Drugs
• Chemical Criminalistics
Forensic biology & DNA laboratory
FB/DNA Laboratory
Evidence Recovery Unit
DNA Laboratory
Casework Laboratory
Person Sample Laboratory
Specialist DNA Laboratory
Case Management
Unit
Database Management
Team
R&D Laboratory
69 FTE employees
DNA laboratory
• Fully-automated DNA laboratory
• Average TAT of ~2 days
21/09/2015
3
Forensic biology & DNA laboratory
• High throughput lab
• In one year:
– 11,660 cases received
– 35,531 samples tested
– 17,680 person samples tested
– 1822 court statements completed
– 8871 database links reported
Impetus for change
Impetus for change
• Trace/touch DNA samples → weak/complex mixtures
• Inconsistency between labs → DNA evidence
suspended in Victorian courts in December 2009
• ANZPAA NIFS standardisation of DNA interpretation
project
• Formation of StatSWG
21/09/2015
4
Impetus for change cont.
• StatSWG recommended move towards continuous
probabilistic model for DNA interpretation
• National standardisation could not be envisaged without a
software solution
• STRmix™ developed to meet needs of Australasian
forensic community
STRmix™
What is STRmix™?
• Fully continuous probabilistic expert system
• Uses Markov Chain Monte Carlo to thoroughly ‘explore’ profile
• Genotype Probability Distribution produced following deconvolution
• Genotype weights used to calculate Likelihood Ratio
• Models stochastic effects, drop-out, drop-in → suitable for weak &
complex profiles
STRmix™ logo used with the permission of the Institute of Environmental Science and Research Limited
21/09/2015
5
Validation
Validation overview
• April – June 2012: beta testing (STRmix™ v1.01, Profiler Plus®)
• January – March 2013: validation (STRmix™ v1.05,
PowerPlex® 21)
• Known samples: single source, 2 & 3 contributor mixtures
• Reproducibility
• Determination of model parameters
Validation overview cont.
March 2013: STRmix™ v1.05 & PP21 implemented for
casework use
• July 2013: Known samples: 4 contributor mixtures
• Ongoing validation & testing when new versions released
• Currently using STRmix™ v2.06 and about to implement
STRmix™ v2.3.06
21/09/2015
6
Validation overview cont.
• STRmix™ v2.3.06 validation → determination of model
parameters, comparison of results with STRmix™ v2.06,
reproducibility
• Future validation → mixture samples with five known
contributors
Key findings
• Rapid → most profiles deconvoluted within several minutes
on a standard PC/laptop
• Accurate → genotypes of known contributors generally
assigned highest weight
• Reproducible
Key findings cont.
• Provides more informative match statistics than existing
binary methods
• Artefacts can significantly effect results if not removed
21/09/2015
7
Training
Training overview
• January 2012: Intro to software and training course
• July 2012: ‘Train the trainer’ course
• August 2012: In-house training of senior reporting biologists
Training overview cont.
• August 2012 – January 2013: theory and practical exams
completed by senior staff, authorised to use STRmix™
• Ongoing training: lectures, discussion of papers
• Lectures to Police, prosecutors & other parties
21/09/2015
8
Challenges
Challenges
• Implementation of new kit & interpretation software was not
without difficulty
• Two of the biggest issues encountered:
– Assigning number of contributors
– Increased workload
Assigning number of contributors
• STRmix™ requires user to assign number of contributors
• Can be difficult to do, particularly for weak/complex
results
• Increased sensitivity of PP21 kit, ABI 3500xL analyser,
extraction chemistry → more mixtures recovered & most
with >2 contributors
• Increased stochastic effects & artefacts such as ‘post
stutter’ & ‘stutter of stutter’ complicate interpretation
21/09/2015
9
Potential solutions
• Analyse using lower analytical threshold, re-test sample
• Deconvolute assuming different number of contributors to
determine effect on LR
• LRs for major contributors generally insensitive if number of
contributors varied
• Can have a significant effect on LRs for minor contributors
however
21/09/2015
10
16,17 15,17 11,19 8,13 16,18
14,15 16,24 10,12 9,139,12
18,20 31.2,33.2 8,10 10,10 8,88,9.3
12,13 17,20 14,15 19,28
Assuming two contributors:16,17
Contrib 1: 89%Contrib 2: 11%
STRmix™ v2.06 results
21/09/2015
11
If 2 contributors assumed, LR = zero (i.e. exclusion)
If 3 contributors assumed, LR approx. 16 billion favouring
inclusion
How to report???
Increased workload
• STRmix™ has greatly expanded our ability to interpret weak
& complex results → increased workload
• Rapid… but can be very time consuming to review results
• Expert system for experts: need to scrutinise output to check
that it ‘makes sense’
• May not model profile appropriately, even for mixtures that
appear fairly straightforward
High stutter (16%)
High stutter (15%)
3rd contributor? Combination ofpost stutter/stutter of stutter?
21/09/2015
12
High stutter (14%)
Contrib 1: 3%Contrib 2: 32%Contrib 3: 65%
STRmix™ v2.06 results
STRmix™ v2.06 results cont.
21/09/2015
13
• Issue reproducible in STRmix™ v2.06 even if a more
thorough sampling of profile carried out
• Improved modelling in STRmix™ v2.3.06 → results align with
intuitive expectations
STRmix™ v2.3.06 results
Contrib 1: 1%Contrib 2: 33%Contrib 3: 66%
STRmix™ v2.3.06 results cont.
21/09/2015
14
Solutions
• Up-skilling of non-reporting biologists
• No further interpretation of complex mixtures in a
case if ‘simpler’ results available to interpret
“Due to the low level and/or complexity of the mixture,
further interpretation has not been carried out.”
Summary
• Expert systems such as STRmix™ enable interpretation of weak &
complex results that would previously have been reported as
‘inconclusive’
• Validation of STRmix™ by NSW FASS and other Australasian
laboratories has demonstrated its effectiveness
• Objective system but does rely on subjective decisions by analyst
e.g. number of contributors, removal of artefacts etc.
• Results need careful review → expert system for use by experts
Questions?zane.kerr@health.nsw.gov.au
top related