NicheVision and ESR(subcontractor) propose to combine our considerable knowhow and experience in probabilistic genotyping (ProbGen) and next generation sequencing (NGS) to develop a commercial solution to the problem of deconvoluting mixed NGS samples using either STR or SNP markers. We propose to develop and validate the models needed to model NGS STR and SNP data by leveraging the ESR experience with modeling CE data in the commercial CE-based STRmix fully continuous NGS-based ProbGen product. Knowhow in ProbGen for CE data will be applied to model the considerably more complex sequence based STR data as well as SNP data. NGS modeling will include statistical characterization of allele read count intensities and variability; as well as stutter intensities as categorized by sequence-based stutter motif pattern in order to associate stutter artifacts with parent alleles and thereby render them recognizable as stutter artifacts rather than drop in alleles. A model for DNA degradation will be built based on data generated for the project and knowhow accumulated for amplicon degradation modeled in CE systems and applied in the current STRmix software. We will apply NicheVisions knowhow used in the MixtureAce binary analysis NGS software for categorizing sequence based STR sequence types into alleles, stutter artifacts and non-stutter artifacts (primarily sequencing error) so that artifacts can be filtered from alleles prior to ProbGen deconvolution of allelic profiles. The interpretation of sequence-based alleles and artifacts represent a paradigm change from current practices used in CE length-based allele number methods and stands as a barrier to practical adoption of NGS technology in forensic laboratories. Therefore, we propose to develop a visualization tool as an interpretation aid for mixed NGS samples that resolves the isometric allele and artifact stacking issue that exists in current NGS software such as Verogens UAS which are designed for single source samples. In addition to sequence-based STR marker deconvolution, this effort will address deconvolution of SNP markers, which when generated by NGS technology can be modeled in many ways as the same thing as STR alleles that show only two possible alleles and that result in excessive allele stacking in mixed samples. NicheVision will apply its knowhow in estimating the number of contributors and the mixture proportion ratios from NGS-SNP data as input to benefit the downstream ProbGen process. At the conclusion of Phase I research, we will deliver a fully functional prototype solution to deconvoluting NGS-based STR and SNP markers using a fully continuous ProbGen approach. Strengths and weaknesses of the prototype(s) will be described in the final report for the purpose of addressing them in the Phase II effort.
