Solving Crimes Using MCMC to Analyze Previously Unusable DNA Evidence

• Activity Number: 60
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistical Computing
• Author(s): Mark Perlin*+
• Companies: Cybergenetics
• Keywords: DNA analysis ; Markov chain Monte Carlo ; Bayesian modeling ; computational forensics ; mixture interpretation ; MATLAB application

Abstract

A crime lab can amplify minute quantities of DNA into highly informative data. With much DNA from one person, the genetic type ("genotype") is evident, and comparing two genotypes gives a match statistic. But a mixture of several people, or having less DNA, introduces genotype uncertainty. Manual data analysis thus understates the match statistic or discards usable evidence as "inconclusive".

Hierarchical Bayesian modeling can statistically describe how labs transform biological evidence into DNA data. Solving these probability equations by MCMC computation separates out the genotypes in a mixture. Data uncertainty translates into genotype probability, later used in a likelihood ratio to yield a match statistic. This objective and thorough analysis preserves identification information.

The MATLAB-based TrueAllele® Casework system implements this MCMC approach. TrueAllele calculates accurate match statistics in rapes and murders when human analysts cannot. The system reanalyzed the World Trade Center DNA data to identify victim remains. By using more of the data, TrueAllele can analyze previously unusable evidence. This talk describes the method, and shows how it solves crimes.