Browsing by Subject "DNA phenotyping"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Advancements in forensic DNA-based identification
    (2017) Dembinski, Gina M.; Picard, Christine; Christie, Mark; Walsh, Susan; Randall, Stephen; Goodpaster, John
    Modern DNA profiling techniques have increased in sensitivity allowing for higher success in producing a DNA profile from limited evidence sources. However, this can lead to the amplification of more DNA profiles that do not get a hit on a suspect or DNA database and more mixture profiles. The work here aims to address or improve these consequences of current DNA profiling techniques. Based on allele-specific PCR and quantitative color measurements, a 24-SNP forensic phenotypic profile (FPP) assay was designed to simultaneously predict eye color, hair color, skin color, and ancestry, with the potential for age marker incorporation. Bayesian Networks (BNs) were built for model predictions based on a U.S sample population of 200 individuals. For discrete pigmentation traits using an ancestry influenced pigmentation prediction model, AUC values were greater than 0.65 for the eye, hair, and skin color categories considered. For ancestry using an all SNPs prediction model, AUC values were greater than 0.88 for the 5 continental ancestry categories considered. Quantitative pigmentation models were also built with prediction output as RGB values; the average amount of error was approximately 7% for eye color, 12% for hair color, and 8% for skin color. A novel sequencing method, methyl-RADseq, was developed to aid in the discovery of candidate age-informative CpG sites to incorporate into the FPP assay. There were 491 candidate CpG sites found that either increased or decreased with age in three forensically relevant xii fluids with greater than 70% correlation: blood, semen, and saliva. The effects of exogenous microbial DNA on human DNA profiles were analyzed by spiking human DNA with differing amounts of microbial DNA using the Promega PowerPlex® 16 HS kit. Although there were no significant effects to human DNA quantitation, two microbial species, B. subtilis and M. smegmatis, amplified an allelic artifact that mimics a true allele (‘5’) at the TPOX locus in all samples tested, interfering with the interpretation of the human profile. Lastly, the number of contributors of theoretically generated 2-, 3-, 4-, 5-, and 6-person mixtures were evaluated via allele counting with the Promega PowerPlex® Fusion 6C system, an amplification kit with the newly expanded core STR loci. Maximum allele count in the number of contributors for 2- and 3-person mixtures was correct in 99.99% of mixtures. It was less accurate in the 4-, 5-, and 6-person mixtures at approximately 90%, 57%, and 8%, respectively. This work provides guidance in addressing some of the limitations of current DNA technologies.
  • Thumbnail Image
    Item
    Evaluation of supervised machine-learning methods for predicting appearance traits from DNA
    (Elsevier, 2021) Katsara, Maria-Alexandra; Branicki, Wojciech; Walsh, Susan; Kayser, Manfred; Nothnagel, Michael; VISAGE Consortium; Biology, School of Science
    The prediction of human externally visible characteristics (EVCs) based solely on DNA information has become an established approach in forensic and anthropological genetics in recent years. While for a large set of EVCs, predictive models have already been established using multinomial logistic regression (MLR), the prediction performances of other possible classification methods have not been thoroughly investigated thus far. Motivated by the question to identify a potential classifier that outperforms these specific trait models, we conducted a systematic comparison between the widely used MLR and three popular machine learning (ML) classifiers, namely support vector machines (SVM), random forest (RF) and artificial neural networks (ANN), that have shown good performance outside EVC prediction. As examples, we used eye, hair and skin color categories as phenotypes and genotypes based on the previously established IrisPlex, HIrisPlex, and HIrisPlex-S DNA markers. We compared and assessed the performances of each of the four methods, complemented by detailed hyperparameter tuning that was applied to some of the methods in order to maximize their performance. Overall, we observed that all four classification methods showed rather similar performance, with no method being substantially superior to the others for any of the traits, although performances varied slightly across the different traits and more so across the trait categories. Hence, based on our findings, none of the ML methods applied here provide any advantage on appearance prediction, at least when it comes to the categorical pigmentation traits and the selected DNA markers used here.