20-Jun-2024

TrueAllele v. Other Software – Choosing data changes answer

Suppose a lawyer could pick a DNA match answer to support their case. Need to convict? Have your expert choose their lab data in a way that gives a large inclusionary match statistic. Need to acquit? Have them select data differently to report a small exclusionary match statistic. Same data, different answers.

Other software can do this. In New York v. Hillary, a higher 50 rfu threshold discarded low-level DNA data, leading STRmix™ software to statistically include the defendant. But a slightly lower 30 rfu threshold retained potentially exculpatory data to statistically exclude him. Fortunately, Cybergenetics assisted Legal Aid pro bono in explaining the science to the judge, and helped acquit an innocent man.

In a June Journal of Forensic Sciences paper “A Tale of Two Systems,” Dr. Susan Greenspoon and her colleagues at Virginia’s Department of Forensic Sciences compared two probabilistic genotyping (PG) systems. Cybergenetics TrueAllele has Bayesian math that uses all the DNA data, without data thresholds or choices, giving reproducible answers. But other software is sensitive to data input. For example, STRmix depends on thresholds and other choices for input. Human data choices can change its LR match statistic.

The above bar chart (adapted from the paper’s Figure 4 underlying data) shows inclusionary STRmix match statistics at a higher 75 rfu threshold with less DNA input (blue). But when STRmix sees more data at a lower 10 rfu cutoff (orange), the log(LR) stat flips and turns exclusionary. These DNA mixtures contained 2 to 4 contributors, including a highly degraded item with only partial data. Human review and TrueAllele gave exclusionary results for the item, based on the observed evidence. STRmix gave conflicting answers, depending on the user's input data choices.

Links

Suspect-centric bias in DNA mixture interpretation - Publications
A tale of two PG systems: A comparison of the two most widely used continuous probabilistic genotyping systems in the United States - Journal of Forensic Sciences