Machine Learning Aided De Novo Design Identifies Novel Benzimidazolone Based Inhibitor‐Modulators for Heat Shock Protein 90 (HSP90)

Abstract

AbstractThe present study investigates the identification of novel heat shock protein‐90 (HSP90) inhibitors‐modulators employing de novo drug design methodologies. Precisely, a machine learning (ML) assisted de novo design tool, REINVENT4, was employed for generating novel molecules against a given set of known HSP90 inhibitors. Further, a series of cheminformatics analyses were undertaken to assess pharmacokinetic properties, evaluate binding affinity using docking, and predict absolute binding affinity using the KDeep tool to reduce the chemical space. Finally, three novel benzimidazole‐based drug‐like candidates viz. IM1, IM2, and IM3 were selected as potent inhibitors‐modulators for HSP90 protein. Docking‐based binding affinity of IM1, IM2, and IM3 was obtained as −11.30, −11.50, and −11.20 kcal/mol, respectively. Moreover, all identified protein‐ligand complex dynamic behavior has been extensively assessed through MD simulation for a 100 ns span, highlighting overall interaction stability and conformational rearrangements for protein bound with all selected compounds. The MM‐GBSA‐based binding free energy estimation for each complex demonstrated the superiority of identified compounds, as found to be −43.54, −38.34, and −25.50 kcal/mol for IM1, IM2, and IM3. Collectively, pharmacokinetics and drug‐like profile assessment of identified compounds, along with simulation trajectories, explained the potentiality of the compounds for modulating the activity of the HSP90.