Weight-of-evidence for DNA identification of missing persons and human remains using CODIS

DNA Identification of unidentified human remains (UHR) is performed in Israel by comparing the UHR's short tandem repeat (STR) profiles to a national database of STR profiles taken from relatives of missing persons. Kinship analysis is performed using the CODIS 7.0 software and results are stated as a Joint Pedigree Likelihood Ratio (JPLR). The weight-of-evidence for JPLR has never been studied, making it difficult to interpret the meaning of specific values in terms of whether UHR are related to specific pedigrees. Therefore, the aim of this study was to statistically determine the practical meaning and context of the JPLR. We used 440 million pairs of simulated DNA profiles and 294 pairs of real ones from known siblings, parent/offspring and unrelated persons. A Score-Based Likelihood Ratio (SBLR) was empirically constructed, validated and compared to both JPLR and the LR produced by CODIS. Our results show that CODIS's JPLR and LR values for single-person pedigrees overestimate the level of support for both "parent/child" and "siblings" propositions relative to the "unrelated" proposition, by up to two orders of magnitude. A practical table is given for correcting this phenomenon, with statistical interpretation (i.e. SBLR) for each JPLR score, including verbal levels of propositional support ranging from "no support" (SBLR<2) to "extremely strong" (SBLR>1 Million).