AI-Driven Discovery of Niclosamide as a VTRNA1-1 Mimic Targeting the p62 ZZ Domain in Acute Myeloid Leukemia

Global transcriptomic profiling via RNA-seq revealed that VTRNA1-1 depletion triggers a systemic collapse of the MYC and FOXM1 regulatory axes. To identify pharmacological agents capable of mimicking this transcriptomic signature, we developed a machine learning pipeline utilizing a Random Forest classifier trained on the L1000FWD database. Virtual screening of Phase 4 approved drugs identified the anthelmintic drug niclosamide as a top-ranked candidate with a 98.17% mimic probability.

Explainable AI (XAI) analysis validated the structural rationale for this prediction, highlighting specific chemical fragments within niclosamide's salicylanilide core. Furthermore, molecular docking simulations predicted that niclosamide binds to the ZZ domain of the p62 (SQSTM1) scaffold protein with favorable binding energy (−7.5 kcal/mol), forming a critical conventional hydrogen bond with GLY 343. These findings suggest that niclosamide acts as a physical "lid" over the p62 ZZ domain, effectively phenocopying VTRNA1-1 loss and suppressing oncogenic signaling.

Our study integrates transcriptomic analysis, AI-driven drug repositioning, and structural biology to propose niclosamide as a potential therapeutic candidate for AML. Importantly, while this framework provides a compelling pharmacological rationale, the docking and machine learning findings represent computational predictions that strictly require future experimental validation.