ASCA-SHAP dual screening of PMI-related corneal ATR-FTIR spectral bands under temperature-dependent variation.

The estimation of postmortem interval (PMI) is a long-standing scientific challenge in forensic pathology, and its accuracy is susceptible to environmental conditions. Environmental temperature can affect processes such as corneal tissue dehydration, water migration, and degradation of biological macromolecules, altering the spectral evolution characteristics of the cornea after death. Therefore, the relationship between PMI and spectral features may change with temperature conditions. This study collected attenuated total reflectance Fourier-transform infrared spectroscopy (ATR-FTIR) spectra of 144 rat corneal homogenate samples under three controlled temperature conditions (4°C, 20°C, and 30°C) and six PMI time points (0, 6, 12, 24, 36, and 48 h). Firstly, partial least squares regression (PLSR) models were established for different temperature groups and combined data to compare the modellability of the PMI-spectral relationship under different temperature conditions; subsequently, variance-synchronized component analysis (ANOVA-simultaneous component analysis, ASCA) was used to decompose spectral variations into PMI, temperature, and PMI × temperature interaction effects, and dual screening was conducted using SHapley Additive exPlanations (SHAP) and other equivalent contribution analyses based on PLSR linear coefficients. The results showed that the single-temperature model was superior model performance to the combined, suggesting that temperature affects the relationship between PMI and spectra. The ASCA results indicated that PMI, temperature, and PMI × temperature interaction effects explained 27.38%, 5.94%, and 21.57% of the total spectral variation, respectively. The higher interaction effect suggests that temperature does not merely introduce constant background differences but also alters the spectral evolution trajectory related to PMI. Based on this, this study constructed a dual-screening strategy integrating ASCA effect decomposition and SHAP contribution analysis to identify interpretable PMI-related spectral bands under temperature-dependent variations. Through this screening framework, 56 wave numbers were identified and merged into 9 consecutive candidate spectral bands. These spectral bands are mainly related to protein, amide region, carbonyl group and fingerprint region molecular changes, suggesting that they may reflect the molecular degradation process of the cornea after death. The above spectral bands exhibit strong PMI-related effects, relatively low temperature main effects, and relatively consistent cross-temperature change trends under the current controlled conditions. However, they should not be interpreted as completely independent biomarkers that are completely unaffected by temperature. The proposed ASCA-SHAP dual screening framework in this study can provide candidate spectral features for cross-temperature and interpretable corneal PMI models.