Comparison of PLSR and PCR Models for Estimating Time Since Deposition of Human Saliva Stains Using ATR-FTIR Spectroscopy Under Simulated Crime Conditions

Saliva is a biological fluid that can be found at various crime scenes and mainly presents two challenges for the forensic analysis: its identification and the estimation of the time since deposition (TSD). In this study, the performance of Partial Least Squares Regression (PLSR) and Principal Component Regression (PCR) models is compared for estimating the TSD of human saliva stains using Attenuated Total Reflectance Fourier-Transform Infrared Spectroscopy (ATR-FTIR). Saliva samples were obtained from eight donors and deposited on various surfaces, exposed to different environmental conditions (indoor and outdoor) and analyzed over a period ranging from 0 to 212 days. The results indicate that PLSR outperforms PCR in terms of Root Mean Squared Error (RMSE) in all cases. This improvement is particularly evident on paper surfaces, where PLSR reduces the RMSE by 10.45 days under indoor conditions and by 13.47 days under outdoor conditions compared to PCR. On woven surfaces, PLSR improves the RMSE by 3.19 days under indoor conditions and by 8.27 days under outdoor conditions compared to PCR. These results highlight the potential of vibrational spectroscopy combined with chemometric methods for the in situ forensic analysis of biological fluids at crime scenes.