Reinforcement Learning-Driven Conversion Designed Minibinders to Small Molecules Targeting Porcine Nectin-1 IgV Domain for Anti-PRV Therapy

Abstract The pseudorabies virus (PRV) is a devastating pathogen in swine with substantial economic impact and emerging zoonotic potential. Here, targeting the porcine Nectin-1 IgV domain, a critical mediator of PRV entry, five minibinders were designed to bind to hydrophobic surface, while B1 achieve > 99% viral inhibition at concentrations > 500 nM with an EC 50 of 1.2 nM in vitro . Capitalizing on the structural insights from B1 , we implemented an artificial intelligence (AI)-driven reinforcement learning framework to convert the minibinder’s pharmacophore into small-molecule inhibitors to block the protein-protein interactions (PPI). This approach yielded SM1 , a compound exhibiting extraordinary potency and a high safety margin (EC 50 = 0.42 pM; CC 50 > 100 µM). In vivo studies demonstrated that both B1 (6 mg/kg, BID) and SM1 (30 mg/kg, BID) significantly reduced brain viral loads by > 99% in infected mice. This study not only identifies promising anti-PRV candidates targeting Nectin-1 but also establishes a generalizable AI-powered pipeline for developing small-molecule PPI inhibitors via minibinder-to-small-molecule conversion.