A synergistic deep learning and machine learning framework for screening heterocyclic compounds against ALDH1A1

Aldehyde dehydrogenase 1A1 (ALDH1A1) has emerged as a promising therapeutic target because of its critical roles in cancer stem cell maintenance and chemoresistance. However, the development of highly selective ALDH1A1 inhibitors remains challenging because of the extensive structural conservation shared with the closely related isoforms ALDH2 and ALDH1A2. In this study, we developed an integrated computer-aided drug design (CADD) and artificial intelligence (AI) framework to systematically identify selective ALDH1A1 inhibitors from a heterocyclic compound library. The multistage virtual screening workflow integrated deep learning-assisted molecular docking, convolutional neural network (CNN)-based scoring, and stringent isoform selectivity filtering. Subsequently, a LightGBM-based classification model was applied to prioritize candidate inhibitors, advancing LDN-27219 and TUG-1375 for dynamic validation. The dynamic stability and binding energetics of the selected protein-ligand complexes were further evaluated using molecular dynamics simulations and molecular mechanics-Poisson-Boltzmann surface area (MM/PBSA) calculations. Collectively, the computational analyses suggest that LDN-27219 exhibits favorable binding characteristics and represents a promising lead candidate for subsequent experimental validation. This integrated AI-CADD framework provides an efficient and reliable strategy for the rapid identification and prioritization of structurally novel, isoform-selective ALDH1A1 inhibitors for future drug discovery efforts.