Abstract

The analysis of forensic DNA samples involves algorithm to calculate the (log) Likelihood Ratios (logLRs) at several loci and summing them to obtain a total log likelihood ratio, using one of several open source algorithms. While variation within any one parameter has been explored, which parameter contributes the most variation has not been addressed. In this study, several algorithms were used to calculate the individual per-locus logLRs and their sum (total logLR) for 5 replicates of the same sample, using different total amounts of DNA, different numbers of contributors (2, 3, or 4) and different percentages of each of 2, 3, or 4 contributors. The results show substantially greater variability among the 5 replicates than is commonly assumed. We show that the percent error is not only larger than expected, but also depends strongly on the total amount of DNA for analysis, the amount of DNA for each specific contributor, and, to a lesser extent, even the program algorithm that was used. These findings have important consequences for interpreting forensic DNA likelihoods, where usually only one replicate (not 5) and only one algorithm is available for analysis.