Toward minimal SNP sets for record-matching with CODIS STR profiles

Genetic record-matching is a technique by which profiles with one set of genetic markers can be queried against databases of profiles with a different set of markers to determine if profiles containing different marker sets trace to the same individual. In forensic genetics, the potential for using genetic record-matching to test single-nucleotide polymorphism (SNP) profiles for genetic matches to short-tandem repeat (STR) profiles could enable development of backward-compatible SNP marker systems to ultimately replace existing forensic STR systems. This study aims to identify minimal SNP sets for achieving record-matching accuracies comparable to those previously observed with tens or hundreds of thousands of SNPs. Using phased SNP-STR reference data in a worldwide panel of individuals, we evaluate record-matching accuracy with SNP sets chosen by each of a variety of SNP selection strategies. When selecting SNPs randomly, ~9000 SNPs are required for achieving record-matching accuracy comparable to that seen with the full SNP set in the "needle-in-haystack" matching scenario, namely 99% of SNP and STR profiles correctly paired with no false-positive identifications in the median accuracy for test sets of size 626 profile pairs. Selecting SNPs based on various thresholds for their minimal minor allele frequency and physical distance to the STR, however, panels of 1800 SNPs, and as few as 900 SNPs, suffice. These results advance toward a potential minimal size for backward-compatible forensic SNP systems that proceed by genetic record-matching.