Adapting the DeepSARM approach for dual-target ligand design

Area of Science:

• Medicinal Chemistry
• Computational Drug Discovery
• Pharmacology

Background:

• The Structure-Activity Relationship (SAR) Matrix (SARM) methodology organizes compound series and visualizes SAR patterns.
• DeepSARM enhances analog design using deep learning and generative modeling, incorporating information from related targets for increased structural novelty.

Purpose of the Study:

• To present the foundational SARM methodology.
• To discuss the adaptation of DeepSARM for designing dual-target compounds in polypharmacology.
• To illustrate a computational proof-of-concept for designing dual-target anti-cancer inhibitors.

Main Methods:

• Development and application of the SARM methodology for compound series extraction and organization.
• Extension of SARM with DeepSARM, integrating deep learning and generative modeling.
• Computational proof-of-concept focusing on dual-target inhibitor design for anti-cancer agents.

Main Results:

• Demonstration of SARM's capability to extract and organize structurally related compound series.
• Adaptation of DeepSARM for novel analog design, considering related targets.
• Successful illustration of DeepSARM for designing candidate dual-target inhibitors against prominent anti-cancer targets.

Conclusions:

• The SARM methodology provides a robust framework for SAR analysis and analog design.
• DeepSARM significantly advances target-based analog design, enabling exploration of novel chemical space.
• DeepSARM shows promise for accelerating polypharmacology-oriented drug discovery, particularly for dual-target agents.