Targeting PRMT5 Across Cancers and AI-Driven Discovery of Candidate Compounds

Abstract

Protein arginine methyltransferase 5 (PRMT5) is a promising anticancer target, yet existing inhibitors shown limited clinical efficacy. Here, we paired multi-omics pan-cancer profiling with a machine learning virtual screening pipeline. Bioinformatics analyses with TIMER3, GEPIA2, UALCAN/CPTAC, the Human Protein Atlas, cBioPortal, STRING and Enrichr revealed widespread of PRMT5 mRNA expression, protein abundance, survival, genetic alteration, immune infiltration and co-expression networks with tumor-specific prognostic impact in GBM, LGG and KIRC. A deep neural network screened about 1.5 million PubChem compounds and deep neural network reached a ROC-AUC = 0.9857 and prioritized three candidates, CID 44345182, CID 2794188 and CID 44345334. Subsequent molecular docking against PRMT5 (PDB ID: 4X61), ADMET filtering and toxicity prediction, and 100 ns molecular dynamics (MD) simulations identified CID 44345182 as the potential hit, similar the docking scores of -9.8 compared with -9.2 kcal/mol for the reference ligand in binding affinity and trajectory stability. These findings clarify the context-dependent roles of PRMT5 and further experiment validations are needed to establish CID 44345182 as a promising candidate.