​​Forensic discrimination of sex from thermally treated human hair using ATR-FTIR spectroscopy and machine learning

Thermal degradation alters hair's molecular structure, influencing its protein, lipid, and disulphide components, which serve as indicators for forensic analysis. The present study investigates the effect of thermal treatment on the chemical composition of human hair and explores the potential of ATR-FTIR spectroscopy combined with machine learning for forensic sex determination. ATR-FTIR spectroscopy was employed to analyze untreated and thermally treated hair strands collected from 50 male and 50 female participants aged 18-30 years. The resulting spectral data were subjected to multivariate analysis using PLS-DA, SVM, and KNN models to classify the samples based on sex and thermal treatment status. Thermal exposure caused distinct alterations in the key spectral bands, especially those associated with proteins (Amide I, II, III), lipids (C-H stretching), and disulfides (S-S stretching), indicating structural denaturation, bond cleavage, and oxidative modifications. Furthermore, the application of multivariate models PLS-DA, SVM, and KNN, on ATR-FTIR spectral data proved highly effective in classifying hair samples by sex and thermal treatment status. All three models achived 100% accuracy, precision, recall and F1-scores, effectively distinguishing between thermally treated and untreated samples by sex In conclusion, ATR-FTIR, coupled with advanced machine learning models, offers a powerful, non-destructive tool for assessing thermal damage, characterising hair composition determining sex, offering significant potential applications in forensic investigations involving burnt hair samples.