This paper reports on the features and benefits of a National Institute of Justice (NIJ) grant project that involved the development and validation of a probabilistic system called “NOCit,” which determines a probability distribution on the number of contributors (NOC) to a DNA sample.
DNA samples recovered from crime scenes often contain at least two contributors. Complex forensic DNA mixture interpretation can be challenging, requiring computational advancements that support its use. Using forensic probabilistic tools to identify a DNA sample’s NOC is critical in computing the weight of evidence for a person of interest. Traditionally, the calculation of the NOC for a forensic short tandem repeat (STR) DNA profile involves evaluating peaks per locus and dividing by two, ratios of alleles/allelic balance at a locus, and review to ensure all loci fit the estimated NOC; however, this method yields only an estimate of the minimum NOC that could explain the mixture rather than the probability of a certain NOC. In addition, there can be variation between analysts, which introduces subjectivity when using this method to determine NOC. In addressing these limitations of current practice, Drs. Catherine Grgicak and Desmond Lun at Rutgers University developed and validated a probabilistic system called NOCit. In addition to accounting for the number of alleles, NOCit incorporates models of peak height, including degradation and differential degradation; forward and reverse stutter; and noise and allelic drop-out. Thus, it is a fully continuous system. The algorithm’s performance was recently evaluated by conducting a large-scale validation on 815 ultraviolet-damaged, inhibited, differentially degraded, or uncompromised laboratory-generated DNA mixture samples that contained up to five contributors. It was determined that NOCit calculates accurate, repeatable, and reliable inferences about the NOC. It significantly outperformed manual methods that rely on filtering the signal.