Integrating diffusion models and molecular modeling for PARP1 inhibitors generation

Area of Science:

• Medicinal Chemistry
• Computational Drug Discovery
• Artificial Intelligence in Drug Design

Background:

• Molecule generation is crucial for drug discovery.
• Deep learning models are increasingly used for designing novel compounds.
• Poly (ADP-ribose) polymerase 1 (PARP1) is a key target in cancer therapy.

Purpose of the Study:

• To develop a novel computational approach for generating potential PARP1 inhibitors.
• To leverage diffusion-based generative models and molecular modeling techniques.
• To identify novel chemical scaffolds with potential PARP1 inhibitory activity.

Main Methods:

• Utilized diffusion models for de novo molecule generation from the ZINC20 database.
• Employed a predictive model to estimate PARP1 inhibitory activity of generated compounds.
• Conducted molecular docking and molecular dynamics simulations for binding affinity assessment.

Main Results:

• Successfully generated novel compounds with predicted PARP1 inhibitory potential.
• Identified promising drug candidates through integrated computational screening.
• Demonstrated the efficacy of the combined generative and simulation approach.

Conclusions:

• The integrated deep learning and molecular modeling strategy shows promise for discovering novel PARP1 inhibitor scaffolds.
• This method supports the development of targeted cancer therapies.
• Further research can build upon this approach for accelerated drug discovery.