Is Evaluation of the (hu)MANid Program Estimation of Sex and Ancestry in a Diverse Adult Population

Is the (hu)MANid program valid in a diverse adult population? Farhi, M.1, Marion, I.2, Atsawasuwan, P.1, Nicholas, C.L.1 1. Department of Orthodontics, University of Illinois Chicago, Chicago, IL 2. Department of Pediatric Dentistry, University of Illinois Chicago, Chicago, IL Objectives: (hu)MANid is a free, web-based application that utilizes a worldwide sample of mandibular morphology and metric data to classify ancestry, age and sex of mandibles in a forensic context. Our study aims to test the reliability of this tool with CBCT data from the diverse adult population presenting to University of Illinois Chicago college of Dentistry. Methods: We retrospectively collected a sample of adults aged 20-46 from a racially and ethnically diverse population from the Chicagoland area. Mandibular samples were extrapolated from CBCT data and were analyzed utilizing 3D Slicer and OsiriX. Measurements included metric data (e.g., chin height, mandibular body height at mental foramen etc.) and morphoscopic measurements (e.g., chin shape, lower border of mandible etc.) The data was then input into the (hu)MANid software and analyzed for accuracy of the output prediction of race and sex. Results: In the complete dataset (n=143), we found that (hu)MANid correctly estimated sex 76% of the time, with greater accuracy for known males (95.83%) than known females (65.26%). This level of accuracy is within the range of what is considered to be of forensic utility. For ancestry estimation, the program particularly struggled to correctly identify individuals self-identified as White/Caucasian (percent correct prediction: 18.64%) and Asian (20% correct). Sex prediction is more accurate than seen in a previous subadult sample, but ancestry prediction remains innacurate. Conclusions: (hu)MANid has forensic utility in predicting sex for individuals who are adults (>20 years of age). It is of limited value for predicting ancestry for either subadults or adults in a contemporary, diverse US sample. This may be due to the limitations of the comparative samples that the program was developed from, and might be improved through increasing the size and diversity of the dataset used to create the model underlying this application.