Deep generative model for therapeutic targets using transcriptomic disease-associated data-USP7 case study.

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Deep learning methods are used for in silico drug design but often overlook biological consequences.
• Current strategies focus only on the chemical properties of generated molecules, ignoring disease context.

Purpose of the Study:

• To develop a targeted molecular generation method incorporating biological information.
• To improve the identification of potential drug candidates by considering disease-specific data.

Main Methods:

• Utilized disease-associated gene expression data to guide molecular generation.
• Developed a framework for targeted compound identification.
• Applied the method to generate USP7 putative inhibitors.

Main Results:

• Generated promising compounds for cancer treatment.
• Over 90% of generated compounds exhibited drug-like properties.
• Identified compounds with essential active groups for target interaction.

Conclusions:

• The proposed method offers a novel and reliable approach for generating new compounds.
• Integrating biological context enhances the relevance of generated drug candidates.
• This framework advances in silico drug design by considering disease dynamics.