Combining ATR-FTIR spectroscopy and DEGs with machine learning for necrophagous insect diapause duration estimation

Forensic entomology plays an important role in estimating the postmortem interval (PMI). However, during cold seasons, necrophagous insects may enter diapause, resulting in developmental arrest. Currently, effective methods for estimating diapause duration are limited, which may compromise the accuracy of PMI estimation based on insect evidence. Here we show that multi-dimensional data fusion of gene expression and infrared spectral features could improve the accuracy of diapause duration estimation. Using the forensically important flesh fly Sarcophaga peregrina as a model, we performed transcriptome sequencing across the diapause period and identified 14 core differentially expressed genes (DEGs) with pronounced temporal dynamics. After qPCR validation, 11 DEGs were retained. ATR-FTIR analysis revealed stage-dependent shifts in macromolecular absorption peak features. Source-level integration of the two data types, followed by partial least-squares dimensionality reduction and support vector regression, produced a model (test R² = 0.922) that substantially outperformed either dataset alone. These findings suggest a potential molecular-spectroscopic strategy that may help address the long-standing challenge of PMI estimation in cold-season forensic cases, and offer a preliminary framework for exploring multi-dimensional data fusion in other complex biological processes.