Modelling the dependence structure of Y-STR haplotypes using graphical models. (November 2018)

• Record Type: Journal Article
• Authors: Andersen, Mikkel Meyer; Curran, James; de Zoete, Jacob; Taylor, Duncan; Buckleton, John
• Is Part Of: Forensic science international. Volume 37(2018)
• Page Start: 29
• Page End: 36
• Publication Date: 2018-11
• Subjects: Y-chromosome -- Forensic genetics -- Lineage markers -- Weight of evidence -- Chow-Liu algorithm -- Discrete Laplace method; Forensic genetics -- Periodicals; Génétique légale -- Périodiques; Forensic genetics; Electronic journals; Periodicals; 614.1
• DOI: 10.1016/j.fsigen.2018.07.014
• Languages: English
• ISSNs: 1872-4973

Abstract

Highlights: Demonstrate using graphical models (Bayesian networks) to estimate Y-STR haplotype frequencies. In particular, the Chow-Liu algorithm is used to construct such models that capture dependencies between Y-STR loci. The Chow-Liu algorithm performs almost as well as the discrete Laplace method. Graphical models can handle e.g. intermediate alleles. Abstract: Many methods have been suggested for evaluating the evidential value of a matching Y-chromosomal DNA profile obtained from a biological stain associated with a crime scene and the Y-chromosomal DNA profile of a suspect. Most of these methods are based on estimating the population frequency of the Y-profile. The common independence assumption between loci for autosomal DNA profiles cannot be used for Y-chromosomal DNA profiles. In this paper we reconsider the problem of population frequency estimation by application of Bayesian networks and the Chow-Liu algorithm to model dependencies between loci. We found that the method based on the Chow-Liu algorithm performs almost as well as the discrete Laplace method. We have also made comparisons to the independence model and we have demonstrated once again that the independence method for Y-profiles cannot be supported.