DNA is now routinely used in criminal investigations and court cases, although DNA samples taken at crime scenes are of varying quality and therefore present challenging problems for their interpretation. An innovative and appealing practical modelling framework for complex forensic genetic problems can be provided by Bayesian networks. In particular, in the presence of complicating features such as missing data on individuals, mutation and mixed trace evidence, these problems become extremely challenging both logically and computationally. Using object-oriented Bayesian networks (OOBNs) it is possible to specify a “construction set” of basic model components, which can then be combined flexibly to model and solve a wide variety of problems such as these. Statistical inference is then made efficiently by probability propagation methods for Bayesian networks. Complex problems in criminal identification, paternity testing, identification in DNA mixtures will be illustrated and solved using Hugin software for inference in Bayesian networks. Bayesian networks have been used to structure and solve cases of forensic identification involving DNA traces that might be mixtures of several DNA profiles. Models for a DNA mixture may utilize both discrete information about the alleles present in the mixture and continuous information on the peak heights. The R-package DNAmixtures exploits Bayesian network techniques for efficient statistical analysis of DNA mixtures. DNAmixtures uses the R-package RHugin, which makes Hugin functionality available within R. This course gives a comprehensive overview of the basic knowledge, modelling capabilities and methodology for the probabilistic evaluation of scientific evidence in forensic genetics applications. It will familiarize participants with the basic concepts of Bayesian networks and object-oriented Bayesian networks. It will illustrate the ways in which uncertainty affects the coherent evaluation of forensic evidence and how this issue can be addressed. Extensions of the basic forensic genetic models to account for uncertain allele frequencies (UAF), identity by descent (IBD), and heterogeneous populations (HET) which all generate dependence among founding genes in forensic genetics problems will also be given. Each topic will be illustrated on casework data, partly through hands-on practical tutorials using Hugin software and the R package DNAmixtures. Finally, we aim to enable participants to recognize potential applications of Bayesian networks in their own field of expertise.
- Dawid, A. P., Mortera J., Pascali V. and van Boxel D. (2002). Probabilistic Expert Systems for Forensic Inference from Genetic Markers. Scandinavian J. of Statistics, 29, 577-595.
- Mortera J., Dawid A. P., Lauritzen S. L. (2003) Probabilistic Expert Systems for DNA Mixture Profiling, Theoretical Population Biology, 63, 191-205.
- Dawid A. P., Mortera J. and Vicard P. (2007) Object-Oriented Bayesian Networks for Complex Forensic DNA Profiling Problems, Forensic Science International, 169, 195-205.
- Green P., Mortera J. (2009). Sensitivity of inferences in forensic genetics to assumptions about founding genes. The Annals of Applied Statistics, 3, pp. 731-763, doi: 10.1214/09-AOAS235
- • R. G. Cowell, T. Graversen, S. Lauritzen, and J. Mortera (2015). Analysis of DNA mixtures with artefacts. Journal of the Royal Statistical Society, Series C. (with discussion), 64, 1–-48. Read before the Royal Statistical Society on 11 June 2014.
- Mortera J. (2016) Statistical evaluation of forensic DNA mixtures from multiple traces. In Topics on Methodological and Applied Statistical Inference. Springer International Publishing. doi:10.1007/978-3-319-44093-4_16
- Green, P. J. and Mortera J. (2016) Paternity testing and other inference about relationships from DNA mixtures, On ArXiv: arxiv.org/abs/1609.09638.