Identifying victim remains using kinship genotype inference

W. Allan, M. W. Perlin, "Identifying victim remains using kinship genotype inference", American Academy of Forensic Sciences 75th Annual Scientific Conference, Orlando, FL, 16-Feb-2023.

Talk

PowerPoint presentation of the conference talk.

Download Handout
Download PowerPoint

Abstract

After attending this presentation, attendees will understand how kinship information is used to construct a missing person’s probabilistic genotype. Comparing this genotype with an unknown evidence genotype can calculate a statistical association.

Following a horrific fiery bus crash in Komatipoort, the South African Police Service (SAPS) recovered victims' bodies that were burned beyond recognition. Identification records did not exist for these rural passengers. Although the police collected DNA from the victim remains, it was not helpful – there were no victim reference profiles on file for comparison. But the police also collected DNA from fifteen people who thought their relative might have been on the bus. The police wanted to use the biological relatives’ DNA to identify the victim remains, but their efforts were unsuccessful.

In 2007, Cybergenetics had previously used their TrueAllele® Technology (TA) to identify victim remains from the World Trade Center Disaster [1]. This computer-based DNA interpretation technology reconstructed missing person kinship genotypes (up to probability). The reconstruction used the DNA of relatives, such as parents, children, and siblings. TA database comparison of these kinship genotypes with remains data helped identity victim remains.

In 2009, SAPS reached out to Cybergenetics for help. The police sent them the STR DNA profile data from 15 victim remains and 15 possible relatives. They provided the assumed biological relationships between relatives and missing people. None of the missing were related to one another; thus each kinship genotype would be constructed using just one family reference. For one reference the biological relationship was unknown, so there was no kinship information.

Child probabilistic genotypes can be inferred when both parents’ genotypes are known (e.g., by using a Punnett square). This genotype is comprised of allele pair possibilities, and their probabilities. When only one parent genotype is available, a computer can use population probability for the other parent’s genotype to reconstruct the child’s genotype. Probabilistic kinship genotypes for missing parents and siblings are constructed similarly.

In the Komatipoort case, the TA analyst input relative profile and kinship information into the TA computer. For the one reference lacking kinship information, the analyst asked the computer to construct the various parent, child and sibling genotypes.

Within minutes, the computer had created the corresponding kinship genotypes. The expected values of their match statistics (Kullback-Leibler divergence) were 2 to 3 ban, as predicted when using 9 loci of STR data.

The analyst compared the 15 kinship genotypes to the 15 victim remains to calculate match statistics. Nine of the 15 remains were statistically associated with a kinship reference (including the one without kinship information). The likelihood ratios ranged from hundreds to tens of thousands.

No family association was found for the other six remains. Perhaps there was no available biological relative. Or, in this rural area, a "relative" might not be biologically related, since children there might be raised by distant relatives honored as "parents".

This presentation demonstrates the real-world utility of kinship inference in disaster victim identification. We show how kinship is used to infer missing persons’ genotypes, even with sparse or unknown biological relationships.

Links

American Academy of Forensic Sciences 75th Annual Scientific Conference - Program