Data-Driven Design of PROTAC Linkers to Improve PROTAC Cell Membrane Permeability.

Area of Science:

• Medicinal Chemistry
• Drug Discovery
• Computational Chemistry

Background:

• Proteolysis-targeting chimeras (PROTACs) are an emerging therapeutic modality for targeted protein degradation.
• Optimizing PROTAC physicochemical properties, especially cell membrane permeability, is critical for their efficacy.
• Current PROTAC linker design lacks explicit methods to enhance cell membrane permeability.

Purpose of the Study:

• To develop an AI-driven linker design method for PROTACs that explicitly optimizes cell membrane permeability.
• To create a generative model capable of designing PROTAC linkers with improved physicochemical properties.

Main Methods:

• Developed PROTAC-TS, a linker generative model integrating a chemical language model and reinforcement learning.
• Constructed and validated a predictive model for cell membrane permeability (R² = 0.710).
• Integrated the permeability prediction model into the generative framework for de novo linker design.

Main Results:

• Successfully designed PROTAC linkers with high predicted cell membrane permeability.
• Ensured designed linkers maintained PROTAC likeness.
• Demonstrated the capability of PROTAC-TS to control and enhance cell membrane permeability.

Conclusions:

• PROTAC-TS represents a significant advancement in rational PROTAC linker design.
• The developed model accelerates the development of PROTACs with improved cell membrane permeability.
• This approach holds promise for overcoming key challenges in PROTAC drug discovery.