Research and Evaluation of the Implementation and Use of Continuous Probabilistic Genotyping Software to Improve the Interpretation of Forensic DNA Mixtures

The combination of sensitive “mega-plex” STR kits, the increased detection ability of Capillary Electrophoresis (CE) instruments, along with the submission of evidence containing trace quantities of DNA have created complex DNA mixtures for interpretation. Laboratories are moving from threshold-based binary methods of interpretation, which waste information in the evidence profile, to probabilistic methods of interpretation using Likelihood Ratios (LRs). Recent surveys of laboratories estimate that half of all DNA testing laboratories in the U.S. are using a Probabilistic Genotyping software for complex DNA mixtures. The most popular software used in the U.S. is STRmix, currently implemented in 63 laboratory systems. By using more information from the crime scene profile, laboratories can examine samples that were previously uninterpretable using a binary approach. More profiles can now be used to search CODIS in no-suspect cases. Fortunately, nearly all laboratories have successfully navigated admissibility hearings at trial to allow the use of Probabilistic Genotyping software in casework. However, the issue of how reliable the results are for these difficult mixtures (in the form of a low LR) continues to be a difficult issue for laboratories. This is due in part to the PCAST report that deemed mixtures with contributors of 20% or less as unreliable. We will recruit laboratories interested in improving: (1) Standardization of reporting and interpretation, (2) Consistency within and between different laboratories, (3) Confidence in reporting and testimony for laboratory staff, and (4) Communication with stakeholders during investigations and trials. These laboratories will be provided mixtures from the publicly available ProvedIt database which yield low LRs from trace donors. These difficult mixtures cause much angst amongst forensic DNA practitioners. We will survey participants to observe how STRmix is used in interpretation and reporting workflows. Next, we will introduce two published software tools that take the output of STRmix and provide context for the LR of the POI compared to the population (DBLR and AdventLR). After training, we will provide the laboratories with a new set of mixtures and evaluate improvement in the four areas. This investigation will yield metrics that can be used to improve the ability to report and testify to low LRs. Finally, we propose to research effective methods to search DNA databases from the results of a STRmix analysis. We will compare methods from current allele based software to directly searching a database of weighed genotypes from STRmix using DBLR software.