Establishing Exclusion Criteria and the Significance of Inclusion: Developing Tools for the Interpretation of Complex DNA Mixtures

Based on submission by the proposer: The interpretation of complex, DNA mixtures continues to be a challenge. DNA mixtures containing greater than 2 contributors cannot be assessed by binary approaches, and interpretation schemes that rely on classifying an allele as either 'present' or 'absent' are prone to error. Thus, in a multi-institutional, interdisciplinary effort among four partners: 1) Boston University School of Medicine, Program in Biomedical Forensic Sciences, 2) Rutgers University-Camden, Center for Computational and Integrative Biology, 3) National University of Ireland-CMaynooth, Hamilton Institute and 4) Massachusetts Institute of Technology, Department of Computer Science and Electrical Engineering, we propose to continue development of a computational approach and software system that examines the likelihood that a person could be detected as a distinguishable contributor to an item of evidence. In prior work, we developed MatchIt, a computational method for calculating p-values that accounts for quantitative peak height information. Our results show that MatchIt significantly enhances capabilities for assessing the strength of DNA evidence compared to existing approaches, particularly for samples with low amounts of template DNA or many contributors. We propose to significantly expand the scope and accessibility of MatchIt by, first, refinement of the statistical model; second, algorithmic improvements; third, incorporation of additional data; and last, development of a user-friendly software implementation. The DNA mixture interpretation research and software tool is submitted under the Forensic DNA area of interest. The first goal of this proposal is to provide 2400, well-defined, DNA autosomal- and mega-plex mixture profiles by varying the amount and ratio of cellular material in samples containing one, two, three, four and five contributors. These samples will be created via laser micro-dissection and processed using typical forensic laboratory procedures, i.e., extraction -> quantification -> amplification -> fragment analysis. The profiles will be made available to the greater community via www.bu.edu/dnamixtures. The second objective is to refine the statistical models that define stutter, allele dropout, baseline noise, allele peak height/area and confirm cross-kit and cross-platform compatibility. The statistical models will provide the means to interpret complex mixtures with low signal-to-noise by direct comparisons, rather than de-convolution schemes. The third objective is to develop a user-friendly software system that utilizes direct comparisons such that the match statistic (i.e., LRs) can be examined for relevance. Thus, if MatchIt computes that the known genotype cannot be reliably discerned from artifact, noise and other possible contributors to an item of evidence, then the known is not (i.e., is excluded as) a distinguishable contributor to that stain. ca/ncf