Bayesian Forensic DNA Mixture Deconvolution With a Novel String Similarity Measure

Abstract: Forensic scientists have great interest in deconvolving profiles in crime scene samples, particularly to test the presence of a courtroom suspect. However, DNA degradation, low material quantity, and mixed signals from multiple contributors are near-ubiquitous challenges. Additionally, crime scene analysis includes the polymerase chain reaction (PCR) for amplification, but this generates artifact sequences that confound the true allelic signal. Forensic scientists worldwide currently use capillary electrophoresis, a length-based sequencing method, but the advent of massively parallel sequencing technology enables data analysis at a much higher resolution, increasing the ability to discriminate between profiles in the sample. This more advanced data type requires new statistical methods to ensure rigorous analysis. In this work we propose a Bayes factor, computed by Markov chain Monte Carlo, that tests whether a person of interest’s DNA was present at a crime scene. The model is able to include known contributors, such as the victim. As part of this methodology, a novel string edit distance was published to measure similarity between alleles and their artifacts generated by PCR. This edit distance accommodates stutter, a common artifact seen after PCR of the FBI’s target genome locations of choice. The Bayes factor estimates produced by the model demonstrate consistent ability to detect the suspect at extremely low proportions in a mixture of contributors, with decisive rejection if testing a suspect not present.

Zoom Link: https://unc.zoom.us/j/92731259806?pwd=eTc3Ylo0SWNFOHV4eERvSjRrUmlKQT09
Meeting ID: 927 3125 9806
Passcode: 100210

Share this Event