Ancestry estimation in forensic anthropology: accuracy of the AncesTrees software in a Brazilian sample

Abstract The objective of this study is to analyse the accuracy and applicability of the AncesTrees software with respect to a set of cranial measurements of a Brazilian sample consisting of 114 identified skulls from two osteological collections, predominantly composed of European (n = 59), African (n = 35), and admixed individuals (n = 20). Twenty-four different craniometric measurements are performed and input to AncesTrees via two algorithms, one of which is used in three configurations, with different ancestral groups integrated in the model. The software exhibits superior performance in the estimation of European individuals, reaching 73% accuracy, compared with 66% in the African individuals. Those individuals classified as admixed produce a variety of ancestral classifications, mainly European. Overall, the most accurate combination of AncesTrees is obtained using ancestralForest with only the European and African groups integrated into the algorithm, where the accuracy reaches 70%. The applicability of this software to a specific population is fragile because of the high admixing load, making it necessary to create a more representative anthropometric database of the Brazilian people.