Mandibular metric-based sex estimation in Asian populations: a novel OPLS-LDA framework.

Accurate sex estimation is of utmost importance in forensic anthropology and bioarchaeology. Nevertheless, traditional methods are frequently influenced by confounding factors and biases specific to certain populations. This study aims to revolutionize sex classification methods for Asian populations through the innovative integration of Orthogonal Partial Least Squares (OPLS) for dimensionality reduction and Linear Discriminant Analysis (LDA) for evaluating mandibular metrics. A collection of dry mandibles, predominantly sourced from the late 19th century to minimize the influence of modern admixture, was analyzed, measuring 16 key parameters. OPLS was utilized to isolate sex-relevant variations, followed by univariate and multivariate LDA to achieve robust classification, with cross-validation performed to ensure reliability. The results indicated cross-validated accuracies ranging from 75.7 to 87.4% for univariate LDA and from 82.5 to 85.4% for multivariate LDA, with maximum mandibular body height identified as the most significant predictor (87.4%). This pilot study establishes a reliable framework for sex estimation based on mandibular parameters in pooled Asian populations, offering significant implications for forensic and bioarchaeological applications. The effectiveness of OPLS in eliminating confounding factors was highlighted, enhancing the robustness of the findings. Future research is recommended to extend these methodologies to other Asian subgroups and develop LDA models for fragmented mandibles, thereby improving the practical utility of sex estimation techniques in forensic contexts. Valuable insights into sexual dimorphism are provided, emphasizing the potential for standardized sexing protocols across diverse Asian populations, which reinforces the importance of reliable methodologies in forensic investigations.