Probabilistic genotyping (PG) software is an invaluable tool for forensic analysts working to resolve complex mixtures, allowing analysts to draw conclusions about samples that were typically reported as inconclusive.
With approximately 100 laboratories in the United States currently using some form of PG software, cases that have been processed using this method have begun making their way through the court system, leaving analysts wondering how to best explain these techniques to judges, attorneys, and juries.
Michael Coble, Associate Professor and the Associate Director of the Center for Human Identification at the University of North Texas Health Science Center and Jo-Anne Bright, Senior Science Leader within the Forensic Biology group at the Institute of Environmental Science and Research Limited, are co-chairing a workshop during ISHI 31 to provide analysts with some direction.
In the "Arguing the Case for DNA Evidence Based on Probabilistic Genotyping" workshop, they will focus on drawing upon data, logic, case decisions, and scientific literature to present and defend DNA evidence based on probabilistic genotyping.
We interviewed Mike and Jo to learn more about the hurdles labs may experience when bringing PG software online, the advantages they’ve seen after implementing PG software, and resources available.
In your experience, how have labs been adapting to the use of probabilistic genotyping software in comparison to previous methods? What hurdles might a lab experience when bringing PG software online?
Someone recently said that labs were jumping into the adoption of probabilistic genotyping. That has not been our experience at all. Labs have taken a very scholarly approach to PG implementation. Some of the challenges labs are facing are how to communicate the resulting likelihood ratios, particularly if they are low level. This is of course not an issue restricted to PG but is new for many labs whose first experience with LRs is with the use of PG software.
Many labs are beginning to feel pressure to come online with PG software. What advice do you have to relieve the pressure that they are facing?
Labs are most likely to be successful with the implementation of PG software if they treat it like a project. By assigning resources, setting milestones, allowing analysts the time to complete the validation work and being realistic about timeframes labs will succeed. There are many labs in the US and internationally who have adopted PG and are using it for casework – over 50 in the US alone. Reach out to these labs as an additional resource. Find out what lessons they learned. As the saying goes, “the early bird gets the worm, but the second mouse gets the cheese.” We have found that early adopters of PG software are very open to providing guidance and lessons learned.
What are some of the advantages that labs have experienced after implementing PG software into their workflows?
Labs are getting much more information from their forensic DNA profiles. As an example, one lab director announced at AAFS that they now have 19 to 34% more interpretable profiles since implementing STRmix in their laboratory.
Validating PG software can appear both daunting and cumbersome. There seems to be confusion about how many single source samples should be run during the validation process. Is there a magic number that you would run? Are there additional validation guidelines that you can share?
JB: My experience is mostly with STRmix and therefore my answer will be relating to that particular PG software. We recommend as part of STRmix implementation that labs test 100 single source profiles to model the peak height variability inherent to the lab system. We know from empirical data that things like DNA profiling kit, PCR cycle number, capillary electrophoresis instrument, and possibly even injection parameters have an effect of peak height variability. If you change any of these, you may have to re-tune your settings and that would require a performance check of your software.
MC: In addition to the single source samples, the laboratory must also run an appropriate number of mixtures to validate their software. Organizations like SWGDAM, the ISFG, and the UK Forensic Science Regulator have published guidance documents that laboratories may find useful in addition to several recent papers in the peer-reviewed literature.
Can you provide any guidelines on when labs should consider re-validating their PG software? For instance, what kinds of changes may occur in a lab that would warrant retesting? Or, if chemistries change, to what extent would re-validations need to be completed?
A new kit or new type of CE instrument would definitely require retesting of the PG software. Changes to kit chemistries would depend on whether peak height variances changed. It may be that a simple performance check could be undertaken to confirm that the changes are not material.
While some analysts are able to attend in-depth training on PG software, they are often asked to train others in the lab. Do any online training or refresher resources exist for analysts to refer to?
In 2019, the FBI partnered with the NIJ FTCOE to create a series of online webinars on PG. There were 8 modules in the series which was pitched to pitched at DNA analysts and legal professionals. We encourage people to check it out at: https://forensiccoe.org/webinar/online-workshop-series-probabilistic-genotyping-of-evidentiary-dna-typing-results/
Let’s switch gears a little bit and discuss testifying on PG methods. What are some of the challenges that analysts are bringing forward? Are there any success stories that you’ve heard that you’d be able to share?
JB: In a 2019 survey of STRmix users, North American labs had reported over 65,000 cases. Analysts for these labs had given evidence in over 1300 court cases. There are literally 1000s of cases where the STRmix evidence has been successfully presented in court. In just one example, in 2018 DNA Labs International used STRmix to interpret mixed DNA profiles in a triple homicide in Jupiter, Florida.
MC: See our answer above regarding how to communicate low LR results. The movement away from CPI and RMP to LRs can be a challenge for laboratories at first. Other recent successes of PG include the exoneration of two men (one in Texas and one in Georgia) who served a combined 27 years in prison for crimes they did not commit.
How would you recommend analysts answer questions about switching from old mixture interpretation methods versus using PG software, such as reworking old cases? Do you have any guidelines for how to address this in court?
PG software provides substantial advantages for interpreting low-level, complex DNA mixtures over older methods. The software makes better use of the data by accounting for dropout rather than excluding the locus which was a common approach using manual methods. Our older methods were not necessarily wrong… they just didn’t make the best use of all the data. Using PG, we now obtain higher LRs for true contributors and lower LRs for non-contributors.
What would you say to some of the smaller labs who are worrying about the possibility of Frye/Daubert hearings? Are there resources available to assist them?
There are a lot of resources available to help. We encourage analysts to reach out to the developers and other experts. There are many transcripts of other admissibility hearings that are available online. No laboratory should feel that they are alone in this process. Many analysts working in the field of forensic DNA testing may have never had to prepare for or endure an admissibility hearing. There is help out there!
You’re both hosting a workshop called “Arguing the Case for DNA Evidence Based on Probabilistic Genotyping” at ISHI 31 this year. What can attendees expect to learn from this workshop? Who would benefit most from attending? Are there any pre-requisites for attendance?
No prerequisites! Everyone is invited. There has been a LOT of information relating to PG appearing recently either in scientific journals, personal blogs, the courts, and the press. We aim to cut through the noise and give the facts. The goal of the workshop is to highlight the key perceived issues with PG and to point attendees in the direction of the best resources so they can determine the facts for themselves.
Are there any other topics related to PG that you’d like to address in this interview or any words of advice for our readers?
Remember that you are not alone in adopting Probabilistic Genotyping and Likelihood Ratios. Yes, it takes time to purchase, validate, write SOPs, train, and implement new software. However, the benefits of PG far outweigh our previous methods that tend to throw away data.
Michael Coble, PhD, is an Associate Professor and the Associate Director of the Center for Human Identification at the University of North Texas Health Science Center in Fort Worth, Texas. Dr. Coble received his Master’s Degree in Forensic Science and his PhD in Genetics from The George Washington University. He is a Fellow of the American Academy of Forensic Sciences and a member of the International Society for Forensic Genetics. He serves as a member of the OSAC Biological Data Interpretation and Reporting Committee and is an invited guest at the Scientific Working Group on DNA Analysis Methods (SWGDAM). He is a co-editor of the Forensic Biology subject area of WIREs Forensic Science journal and is a member of the editorial board of Forensic Science International: Genetics.
Jo-Anne Bright is a Senior Science Leader at the Institute of Environmental Science and Research Limited (ESR) New Zealand, where she has worked since 1999. She completed her Bachelor of Science degree at Victoria University of Wellington in 1998 and Master of Science degree in Forensic Science at the University of Auckland in 2000. In 2015, Jo was awarded her PhD after studying advanced DNA profile interpretation topics. Jo has co-authored numerous papers and presented at a number workshops in this area, including in the US, UK, Asia, and Europe. Jo is one of the co developers of STRmix™, expert software for the interpretation of forensic DNA profiles.