DeepScaffold: A Comprehensive Tool for Scaffold-Based De Novo Drug Discovery Using Deep Learning

Area of Science:

• Medicinal Chemistry
• Computational Drug Design

Background:

• Drug design aims to discover novel compounds with optimal pharmacological properties.
• Scaffold-based molecular design is an effective strategy for identifying potential drug candidates.

Purpose of the Study:

• To propose a scaffold-based molecular generative model for drug discovery.
• To enable molecule generation based on diverse scaffold definitions and side-chain properties.

Main Methods:

• Developed a generative model for molecule design.
• Incorporated various scaffold definitions (Bemis-Murcko, cyclic skeletons, side-chain specified).
• Model learns chemical rules for adding atoms and bonds to scaffolds.

Main Results:

• Generated compounds were evaluated using molecular docking against DRD2 targets.
• The model demonstrated effectiveness in generating compounds with specific scaffolds.
• Successfully applied to de novo drug design for potential drug candidates with targeted docking scores.

Conclusions:

• The proposed scaffold-based generative model is effective for drug discovery.
• This approach facilitates the design of novel compounds with desired scaffolds and properties.
• The model shows promise for addressing various drug design challenges.