REGENE-NET: AN ENTROPY-WEIGHTED TRUST-DRIVEN FRAMEWORK FOR RELIABLE PROTEIN–PROTEIN INTERACTION RECONSTRUCTION AND GENE PRIORITIZATION

Abstract

Protein–protein interaction (PPI) networks are crucial for understanding the molecular basis of cellular processes and disease mechanisms. However, experimental datasets often contain noisy or incomplete information, leading to unreliable interaction mappings. To overcome these limitations, this study introduces ReGene-Net (Reliable Gene Interaction Reconstruction Network), a novel computational framework that integrates heterogeneous biological data using entropy-weighted fusion and trust-propagation mechanisms. The framework computes source reliability through entropy-based weighting, refines network connections via trust-weighted random walks, and generates biologically coherent PPI maps. Experimental validation across multiple benchmark datasets demonstrates that ReGene-Net achieves superior performance with an accuracy of 96.4%, precision of 93.6%, recall of 94.5%, and F1-score of 94.0%, outperforming existing models such as DeepPPI, TrustNet, and Random Walk. The reconstructed networks exhibit high biological relevance and structural consistency, effectively identifying key hub genes such as RPL5 and HIST1H1E. These results confirm that ReGene-Net is a robust and interpretable tool for PPI reconstruction, disease gene prioritization, and biomarker discovery in complex biological systems.