Small Shifts, Big Impacts: Using Probabilistic Genotyping Software in Forensic Genetics

Abstract

The genetic analysis of crime scene samples is increasingly challenging due to the low DNA quantity and/or quality and associated stochastic effects, that make them highly complex and difficult to analyze. Indeed, technological advances regarding the ability to analyze vestigial samples have been making the weight of the evidence dependent on computational tools, pushing the corresponding statistical analyses beyond the traditional algebraic framework. The standard problem relies on the quantification of the likelihood of a person of interest (a suspect, e.g.) being a contributor to a problem sample recovered from a crime scene, usually a mixture of an unknown number of contributors [1]. Probabilistic genotyping software (PGS) were then developed to deal with this increasing complexity of analysis, allowing the quantification of the weight of the evidence considering several parameters of interest [2-4]. Each laboratory must estimate and establish parameter values specific to their conditions and according to the PGS specifications. Some of the authors of this work presently coordinate an international working commission from Spanish and Portuguese Speaking Working Group of the International Society for Forensic Genetics (GHEP-ISFG). This working commission aims to understand the state-of-the-art regarding the knowledge, use, and implementation of non-binary tools in the forensic genetic routine. Evaluation of the approaches and methodologies used for the statistical interpretation of forensic problems involving DNA mixtures in different worldwide laboratories will be also carried out.
This work aims to evaluate the impact on the quantification of the evidence when different parameters’ conditions and/or models are considered for population coancestry, number of estimated contributors, and analytical factors such as detection threshold, stutters, and drop-in/drop-out. To reach this goal, pairs of real casework samples (composed of a mixture with two or three estimated contributors and a single-source sample associated) were analyzed using three PGS based on different statistical models. Whenever applicable simulations and an algebraic approach were also computed.
The results obtained show that the estimation of these parameters must not be overlooked, since it may considerably impact the outcome. This reinforces the relevance of proper parametrization in the analysis of forensic genetics identification problems, and the importance of practitioners understanding the statistical models of the distinct computational approaches to use them accurately.