Comprehensive Assessment and Benchmark of Deep Generative Models for Proteolysis TArgeting Chimera (PROTAC) Design.

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Proteolysis TArgeting Chimera (PROTAC) technology is a powerful strategy for targeted protein degradation.
• Deep generative models are increasingly used in drug discovery, including PROTAC design.

Purpose of the Study:

• To quantitatively assess the performance of deep generative models for PROTAC design.
• To provide a comprehensive overview of different generative models applicable to PROTAC design.
• To guide researchers in applying these models effectively.

Main Methods:

• Overview of PROTAC-specific and general linker design deep generative models.
• Quantitative evaluation of representative models on a benchmark of 40 experimental protein-ligand complex structures.
• Comparison with a general de novo design model as an ablated baseline.

Main Results:

• The study evaluates the features and generative performance of various molecular generative models.
• A quantitative benchmark assesses model effectiveness for PROTAC design.
• Comparative analysis highlights strengths and weaknesses of different model types.

Conclusions:

• This work provides a critical quantitative assessment of deep generative models for PROTAC design.
• The findings help researchers understand and select appropriate models for targeted protein degradation strategies.
• The study facilitates the practical application of advanced computational methods in PROTAC development.