Back to Presentations

Forensic DNA Inference

M.W. Perlin, J.B. Kadane and R.W. Cotton, "Forensic DNA inference", Seventh International Conference on Forensic Inference and Statistics, Lausanne, Switzerland, 21-Aug-2008.

Downloads

PowerPoint presentation and handout for the International Conference on Forensic Inference and Statistics 2008 talk.

Download Handout
Download PowerPoint
Download Paper

Abstract

DNA evidence is a powerful casework approach to solving crimes. Biological specimens are collected from the scene of crime, from which scientists generate STR laboratory data and then infer DNA genotypes from the STR data peaks. By comparing these genotypes against the genotypes of possible suspects (including DNA databases), matches between genotypes can help identify the culprit.

However, not all groups compare inferred genotypes against genotypes. Many laboratories have forensic scientists use the STR data peaks directly in their match comparison. Others use a likelihood ratio (LR) to compare alternative forensic hypotheses about the peak data, but without first inferring the genotype.

A likelihood ratio inference from DNA evidence assesses two competing hypotheses relative to the data. Typically, the likelihood data is comprised of the original STR peaks obtained from biological specimens. However, it is also possible to divide the DNA inference problem into two separate stages: first obtaining genotypes from the evidence as posterior probability distributions, and subsequently determining the strength of match using a general likelihood ratio statistic.

To implement this two phase DNA inference approach, we describe a likelihood ratio statistic for inferring match strength that can accommodate uncertain genotypes. This LR considers as evidence (not the original STR data, but instead) the event that a match has occurred, under competing hypotheses about who contributed to the samples. The previously inferred genotype probability distributions serve as background information.

Separating genotype inference from match comparison offers significant advantages. These include (i) performing objective, unbiased comparisons, (ii) using highly informative genotypes inferred from statistical computing, (iii) separating crime scene and suspect DNA workflow processes, (iv) enabling sophisticated computer matching on genotype databases, and (v) providing a general framework for current forensic match approaches.