UC Davis

Since its initial development in 1985, forensic DNA analysis has become increasingly important in casework evidence analysis. Forensic DNA typing is widely carried out today using sophisticated instrumentation and highly sensitive reagents. These developments have proven very beneficial to crime laboratories in solving criminal cases. However, an increase in forensic typing sensitivity can also lead to problems in interpreting the resulting profiles. DNA casework analysts are often confronted with complicated results including mixtures, degraded DNA, and/or low copy number DNA all of which are particularly difficult to interpret, deconvolute, and evaluate statistically. To make matters worse, there is no “one best way” to interpret challenging profiles that is agreed upon by the forensics community at large, and many labs rely on manual deconvolution techniques prone to human error. A significant advancement in the analysis of forensic DNA profiles has been the development of complex computer algorithms for mixture deconvolution and/or statistical analysis in a largely automated fashion. Probabilistic genotyping (PG) software has come to the forefront of forensic DNA interpretation and is being used by more and more crime laboratories throughout the world today. There is now a wide variety of forensic DNA analysis programs available. However, no standardization exists, and crime labs may be unsure as to which program they should validate for use in casework analysis. Two popular DNA interpretation tools include the PG software programs, TrueAllele® and STRmix™. These programs use the Markov Chain Monte Carlo method (MCMC) to examine virtually every possible genotype contained in a DNA profile and to provide a statistical value as to the likelihood of each possible profile. Both programs process complicated mixtures more efficiently than manual binary methods, increasing the chances that the findings are robust, reproducible, and admissible in court. TrueAllele® and STRmix™ have both been validated by multiple crime laboratories and used in casework. However, in a court case, NY v Hillary (2016), STRmix™ excluded Hillary as a possible contributor, while TrueAllele® generated an inconclusive result after analyzing the same DNA profile. While TrueAllele® and STRmix™ both use the MCMC method, the difference in final outcomes suggests the programs may differ in sensitivity etc. Such a discrepancy, therefore, calls for a comparison study to better understand the programs, and to assist laboratories in making the best choice for their casework. This study consists of a comparative MCMC analysis of thirty-six mixture sample profiles (Globalfiler®) that included from two to five contributors using TrueAllele® and STRmix™. These mixtures were originally processed using TrueAllele® as part of the Kern Regional Crime Laboratory’s software validation study. The sample files were processed here using two versions of STRmix™, version 2.5 and 2.6. Although some individual interpretation requests produced different MCMC statistics across the three programs, no statistically significant differences were identified in this study. Rather, both the coefficient of determination and Analysis of Variance (ANOVA) statistic showed that the overall data were comparable between software programs when reporting mixture weight contributions and likelihood ratios (LRs) for mixture types of two to five contributors. Any significant differences were investigated and are discussed in detail.