Integrating diffusion models and molecular modeling for PARP1 inhibitors generation

Molecule generation is a critical task in drug discovery, with growing interest in using deep learning to design new compounds. In this study, we propose a novel approach to generate potential PARP1 inhibitors by combining diffusion-based generative models with molecular modeling techniques. Starting from the ZINC20 database, we used diffusion models to create new compounds and applied a predictive model to estimate their PARP1 inhibitory activity. Promising candidates were further evaluated using molecular docking and molecular dynamics simulations to assess their binding affinity. Our results demonstrate the potential of this integrated method to discover novel scaffolds for PARP1 inhibition, supporting future efforts in targeted cancer therapy development.