Date of Graduation
Eberly College of Arts and Sciences
Forensic and Investigative Science
Probabilistic genotyping systems use sophisticated algorithms to aid DNA analysts with complex mixture interpretation. These systems are used to generate a likelihood ration (LR) which is used to provide the analyst with the statistical weight of the match between an evidence profile and a known contributor. Complex mixtures occur from samples consisting of more than two contributors and samples that have experienced degradation or consist of low template quantity.
This study was performed to compare the performance of three mixture analysis software tools on the same set of samples to determine the capabilities of each software, false inclusions, and false exclusions. Three probabilistic genotyping systems were evaluated: GeneMapper ID-X™ v1.4, NOCIt/CEESIt, and TrueAllele® using 750 samples provided from the PROVEDIt database. The sample set consisted of 100 single-source samples and 650 samples containing 2-,3-,4-, and 5-person samples with template quantities ranging from 0.016 to 1 ng of DNA. Each software was evaluated using the recommended default settings. Each sample was analyzed using the true number of contributors and determined the LR value for each of the contributors present in the sample. To determine the number of false inclusions each sample was analyzed using two incorrect individuals. NOCIt performed better than GeneMapper ID-X™ in regard to determining the number of contributors. However, the NOCIt tool despite being more accurate than GeneMapper ID-X™ experienced a decrease in accuracy as the number of contributors increased. TrueAllele® reported the fewest number of false inclusions across one to five contributor sample sets, but it also reported a higher number of false exclusions when compared to CEESIt. GeneMapper ID-X™ is only able to compute LR values for 1- and 2-person samples, and it will only produce the LR if the software determines that the selected individual is present within the sample which caused it to report the highest number of false exclusions. There was no statistical difference in the LR values obtained from CEESIt and GeneMapper ID-X, but CEESIt and GeneMapper ID-X™ did produce statistically higher LR values when compared to TrueAllele® for the 1- and 2-person samples. With the exception of the three contributor samples CEESIt did not produce statistically higher LR values than TrueAllele®.
This study shows the value of comparing different probabilistic genotyping systems to determine which software tool is the most accurate for casework samples. Future studies should be performed to determine the optimal settings for each software and compared to other systems on the market such as STRMix™. This study shows the value of using statistical algorithms to aid in mixture sample deconvolution to help improve accuracy.
Newland, Kristen, "Evaluation of the Performance of Probabilistic Genotyping Software on Complex Mixture Samples" (2020). Graduate Theses, Dissertations, and Problem Reports. 7720.