Automated familial search using a probabilistic genotype database

M. M. Legler, K. W. P. Miller, M. W. Perlin, G. Sugimoto, "Automated familial search using a probabilistic genotype database", American Academy of Forensic Sciences 74th Annual Scientific Conference, Seattle, WA, 25-Feb-2022.

Talk

PowerPoint presentation of the conference talk.

Download Handout
Download PowerPoint

Abstract

Familial searching can find suspects by searching a DNA database that contains their relatives. Genotypes from people who contributed their DNA to crime scene evidence are compared with reference profiles in the DNA database. Even when the contributor genotypes are not in the database, genetic similarities to contributor relatives who are in the database can help find investigative leads.

Earlier familial search strategies were based upon partial allele matches and simple allele inclusion statistics. Cybergenetics TrueAllele® technology provides a more sophisticated approach using information-preserving probabilistic genotypes, and likelihood ratio (LR) information.

The TrueAllele database automatically generates probabilistic familial genotypes from reference genotypes, using them for automated comparison with probabilistic genotypes from complex evidence. LR match statistics are calculated automatically. This general computer method allows any DNA evidence (e.g., mixtures of up to ten contributors) to be easily compared with automatically inferred familial genotypes.

The Kern Regional Crime Laboratory (KRCL) conducted three validation studies prior to deploying their information driven automated TrueAllele familial search system.

An Accuracy Study demonstrated the software can provide informative match statistics to evidence items when compared to inferred kinship genotypes from relatives of true contributors. DNA profiles from 26 individuals from four different family trees were typed and included in the accuracy study. There were 676 pairwise comparisons made for each assumed genotype group (PCH, SIB and HAS) for a total of more than 2,000 comparisons. The evidence to parent/child (EVI-PCH) and evidence to sibling (EVI-SIB) searches were shown to be informative.

A Sensitivity and Specificity Study examined LR comparison of PCH and SIB inferred genotypes when the mixture weight of the true contributor’s DNA varied. Thirty-one single source profiles and 36 mixture samples (2 to 5 contributors) were compared to the inferred genotypes of 44 single source known reference profiles. The EVI-PCH searches were more specific for inferred genotypes of closely related relatives of known contributors. The EVI-SIB searches were more sensitive for inferred genotypes of more distant relatives of known contributors. The EVI-PCH false positive rate was ~0.02% (1/6512), with log(LR) = 2.117 ban. The EVI-SIB false positive rate was ~0.25% (16/6512), having log(LR) values from ~2.0 to 4.5 ban.

A Known and Nonprobative Evidence Study compared kinship inferred genotypes in a larger database comparable to the KRCL's local DNA database. This study included known reference samples from known matches, used to compare to EVI profiles in the laboratory’s local database of over 5,000 profiles. The EVI-PCH and EVI-SIB searches accurately matched to the inferred genotypes of known contributors, as well as the inferred genotypes of the true contributor’s son. The EVI-PCH comparisons were more specific than EVI-SIB, resulting in fewer false positives. EVI-PCH search did not result in any match scores above 1,000, while EVI-SIB search gave 25 false positives with match scores greater than 1,000. The false positive rate (for a log(LR) ≥ 2) was ~.0001% for EVI-PCH searches, and ~.0006% for EVI-SIB searches.

Familial search is a powerful tool for finding investigative leads from DNA evidence when other avenues have not produced suspects. The KRCL has conducted validation studies and developed protocols to bring automated TrueAllele familial searching into routine forensic practice. This forensic science advance can make more identification information available to criminal justice by reusing data from existing DNA evidence.

Links

American Academy of Forensic Sciences 74th Annual Scientific Conference - Program