Conceptual

GAN-TAT: Graph-Embedding of Protein Interaction Networks for Druggable Gene Identification

A computational framework (GAN-TAT) that identifies druggable genes by directly embedding Protein Interaction Networks (PINs) rather than integrating them indirectly. Because PINs are high-dimensional and sparse, GAN-TAT uses ImGAGN, a GAN-based graph embedding method designed for imbalanced networks, to learn node representations that feed a druggability classifier. On three Pharos datasets it achieves a top AUC-ROC of 0.951 on the Tclin class, with predictions corroborated by clinical evidence, demonstrating the value of directly exploiting network topology for drug-target discovery and pharmacogenomics.