BInD: Bond and Interaction-Generating Diffusion Model for Multi-Objective Structure-Based Drug Design.

Area of Science:

• Computational chemistry
• Structural biology
• Artificial intelligence in drug discovery

Background:

• Advancements in geometric deep generative models and structural data enable structure-based drug design (SBDD) using only target protein information.
• Existing models often struggle to balance multiple objectives, leading to suboptimal performance in specific tasks.

Purpose of the Study:

• To introduce BInD, a novel diffusion model designed to co-generate molecules and their interactions with target proteins.
• To address the limitations of existing models in balancing multiple objectives for SBDD.
• To enhance target binding and specificity through molecule design and optimization.

Main Methods:

• Development of BInD, a diffusion model incorporating knowledge-based guidance for co-generation.
• Ensuring balanced consideration of target-specific interactions, molecular properties, and local geometry.
• Implementation of an NCI-driven molecule design and optimization strategy.

Main Results:

• BInD demonstrates robust performance across all key objectives, matching or surpassing state-of-the-art methods.
• The model successfully co-generates molecules with balanced properties and target interactions.
• The proposed optimization method enhances target binding and specificity.

Conclusions:

• BInD offers a balanced approach to SBDD by co-generating molecules and their interactions.
• The model represents a significant advancement in generative models for drug discovery.
• The NCI-driven optimization further refines molecule design for improved therapeutic potential.