Forensic DNA laboratories in the United States have undergone a transformational shift in how they interpret and report DNA mixtures. Starting in 2010, the Scientific Working Group on DNA Analysis Methods (SWGDAM) recommended the use of a stochastic threshold to establish a stochastic threshold (e.g. 150 rfu) where allelic drop out was possible when peaks below this threshold were observed in a DNA profile. For the most part, laboratories would present the statistical assessment of the evidence profile using a modified Random Match Probability (mRMP) or the Combined Probability of Inclusion (CPI). This approach is often referred to as the “Binary” method of DNA mixture interpretation, where all acceptable genotypes (RMP) or alleles (CPI) have equivalent impact (equal weights) on the calculation.
Over the last decade, probabilistic methods of complex DNA mixture interpretation have been embraced and adopted by the forensic DNA community. Probabilistic Genotyping (PG) Systems use molecular modeling and statistical algorithms to provide weight to genotype combinations to best explain the evidence. In short, better fitting genotypes have a higher weight, affecting the assigned likelihood ratio (LR) than poorer fitting options. Unlike the binary method of interpretation, PG does not use a stochastic threshold. Rather, the software models and evaluates the evidence probabilistically. The statistical strength of the evidence is presented as a Likelihood Ratio (LR), which is a component of Bayes’ Theorem.
In the move from binary to probabilistic methods of interpretation, forensic DNA analysts have had to become familiar with the use of the LR which is not a probability in the traditional sense of a mRMP or CPI. Instead, the LR is a ratio of two conditional probabilities which represents the strength (or weight) of the evidence given two competing hypotheses (e.g. the Person of Interest (POI) is a contributor to the evidence profile) versus an alternative hypothesis (e.g. the POI is not a contributor to the evidence profile.
One PG software is STRmix (ESR, New Zealand) and is by far the most widely used software among forensic DNA laboratories in the U.S. We have found that many laboratories in the U.S. seem hesitant to report low LRs, even though they may have reported RMP or CPI statistics of similar magnitude. Many of these laboratories have developed an LR reporting threshold where LRs that fall in the range of 1/1000 (0.001) to 1000 are reported as “inconclusive.” We have observed this range cover LRs from 1/10,000 (0.0001) to 10,000.
The major goals of this project aimed to expand the scientific knowledge for analyzing complex DNA mixtures and address current challenges faced by the analyst in court. We wanted to research and evaluate methods to help increase the confidence needed for analysts to report low LRs from complex and challenging DNA mixtures. We focused on the PG software STRmix, but feel the same concepts can be applied to other continuous PG software programs that generate weighted genotype probabilities as well. Two software programs, AdventLR and DBLR, present a visual analysis of non-contributor testing which can assist the analyst to convey the strength of low LRs to the trier of fact.
(Author abstract provided.)