Greedy regression-based cross-species gene expression modeling: a translational framework for forensic wound age estimation.

Forensic wound age estimation faces a significant translational gap, as molecular findings from indispensable animal models are challenging to apply to human casework due to the lack of systematic cross-species frameworks. To bridge this gap, we developed a computational framework for cross-species gene expression prediction and injury time inference. Using skeletal muscle contusion models in rats and pigs as a proof-of-concept, we employed greedy regression on 6,327 orthologous genes to construct a total of 3,058 translation models, comprising 284 regression equations based on genes with parallel expression trends across species and 2,774 equations based on conserved expression ratios. The framework successfully predicted pig gene expression from rat data, and a refined set of 156 high-confidence predicted genes enabled an LDA classifier to achieve 93.3% accuracy in cross-validation and 81.1% accuracy on independent porcine samples for injury time classification. This study establishes a generalizable computational paradigm that, by demonstrating robust cross-species transferability in a rat-to-pig system with human-relevant orthologs, provides a methodological proof-of-concept for translating molecular timing information to human forensic casework when experimental data are constrained.