The application of statistical modelling to the interpretation of complex DNA profiles

Abstract

In forensic DNA analysis a profile is typically produced from a biological sample collected from the scene of a crime and compared with the DNA of one or more persons of interest (POI). Single source pristine profiles are relatively easy to interpret and their analysis has achieved worldwide acceptance as a reliable scientific method. However, profiles from crime scenes are frequently compromised in quality, or quantity or both. Stochastic factors are often present in compromised profiles which complicate interpretation. Stochastic factors can include; heterozygous balance, allelic dropout, and increased stutter peaks and are characteristic of low template DNA (LtDNA) samples. Complicating interpretation even further is that in many cases, crime scene samples are composed from two or more people. The number of contributors can be unclear. The presence of three or more alleles at any locus signals the existence of more than one contributor, although it can be difficult to distinguish between the presence of a low level second contributor and stochastic effects. This research investigates the behaviour of LtDNA profiles. The traditional guidelines, used in conventional DNA interpretation, are investigated with respect to their application to LtDNA profiles. Statistical models are created for; heterozygous balance, dropout, and stutter. These models use the explanatory variables identified in the data exploratory section of this research to describe the behaviour of each of the aforementioned stochastic effects. These models have been built with the aim that they will be implemented in a probabilistic approach to DNA interpretation. This research also examines different approaches for the interpretation of forensic DNA profiles and the available methods for the calculation of the weight of the evidence of the profile. A two person LtDNA mixture is interpreted using different methods and the resulting statistics are presented. This is done with the aim of demonstrating how the same LtDNA evidence can be interpreted differently under current interpretation guidelines. It is important that that limits of current interpretation models are understood and are not extended beyond their means.