Abstract
In this study, we leveraged a sophisticated active learning model to enhance virtual screening for SQLE inhibitors. The model's improved predictive accuracy identified compounds with significant advantages in binding affinity and thermodynamic stability. Detailed analyses, including molecular dynamics simulations and ADMET profiling, were conducted, particularly focusing on compounds CMNPD11566 and its derivative HCJ007. CMNPD11566 showed stable interactions with SQLE, while HCJ007 exhibited improved binding stability and more frequent interactions with key residues, indicating enhanced dynamic adaptability and overall binding effectiveness. ADMET data comparison highlighted HCJ007s superior profile in terms of lower toxicity and better drug-likeness. Our findings suggest HCJ007 as a promising candidate for SQLE inhibition, with significant improvements over CMNPD11566 in various pharmacokinetic and safety parameters. The study underscores the efficacy of computational models in drug discovery and the importance of comprehensive preclinical evaluations.