Academic

FuseDiff: Symmetry-Preserving Joint Diffusion for Dual-Target Structure-Based Drug Design

arXiv:2603.05567v1 Announce Type: new Abstract: Dual-target structure-based drug design aims to generate a single ligand together with two pocket-specific binding poses, each compatible with a corresponding target pocket, enabling polypharmacological therapies with improved efficacy and reduced resistance. Existing approaches typically rely on staged pipelines, which either decouple the two poses via conditional-independence assumptions or enforce overly rigid correlations, and therefore fail to jointly generate two target-specific binding modes. To address this, we propose FuseDiff, an end-to-end diffusion model that jointly generates a ligand molecular graph and two pocket-specific binding poses conditioned on both pockets. FuseDiff features a message-passing backbone with Dual-target Local Context Fusion (DLCF), which fuses each ligand atom's local context from both pockets to enable expressive joint modeling while preserving the desired symmetries. Together with explicit bond gene

J
Jianliang Wu, Anjie Qiao, Zhen Wang, Zhewei Wei, Sheng Chen · · 1 min read · 17 views
#cs.LG

arXiv:2603.05567v1 Announce Type: new Abstract: Dual-target structure-based drug design aims to generate a single ligand together with two pocket-specific binding poses, each compatible with a corresponding target pocket, enabling polypharmacological therapies with improved efficacy and reduced resistance. Existing approaches typically rely on staged pipelines, which either decouple the two poses via conditional-independence assumptions or enforce overly rigid correlations, and therefore fail to jointly generate two target-specific binding modes. To address this, we propose FuseDiff, an end-to-end diffusion model that jointly generates a ligand molecular graph and two pocket-specific binding poses conditioned on both pockets. FuseDiff features a message-passing backbone with Dual-target Local Context Fusion (DLCF), which fuses each ligand atom's local context from both pockets to enable expressive joint modeling while preserving the desired symmetries. Together with explicit bond generation, FuseDiff enforces topological consistency across the two poses under a shared graph while allowing target-specific geometric adaptation in each pocket. To support principled training and evaluation, we derive a dual-target training set and use an independent held-out test set for evaluation. Experiments on the benchmark and a real-world dual-target system show that FuseDiff achieves state-of-the-art docking performance and enables the first systematic assessment of dual-target pose quality prior to docking-based pose search.

Executive Summary

FuseDiff is a novel end-to-end diffusion model that addresses the challenge of dual-target structure-based drug design by jointly generating a ligand molecular graph and two pocket-specific binding poses. The model features a message-passing backbone with Dual-target Local Context Fusion (DLCF) that enables expressive joint modeling while preserving symmetries. FuseDiff achieves state-of-the-art docking performance and enables systematic assessment of dual-target pose quality prior to docking-based pose search. This breakthrough has significant implications for the development of polypharmacological therapies with improved efficacy and reduced resistance. The model's ability to jointly generate binding modes for two target pockets under a shared graph structure represents a major advancement in drug design.

Key Points

  • ▸ FuseDiff is an end-to-end diffusion model for dual-target structure-based drug design
  • ▸ The model jointly generates a ligand molecular graph and two pocket-specific binding poses
  • ▸ FuseDiff features Dual-target Local Context Fusion (DLCF) for expressive joint modeling and symmetry preservation

Merits

Strength in Expressive Joint Modeling

FuseDiff's use of DLCF enables the model to capture complex interactions between the ligand and two target pockets, leading to improved docking performance and pose quality assessment.

Advancement in Polypharmacological Therapies

FuseDiff's ability to jointly generate binding modes for two target pockets under a shared graph structure represents a major advancement in the development of polypharmacological therapies with improved efficacy and reduced resistance.

Demerits

Limited Generalizability to Other Drug Design Tasks

FuseDiff is specifically designed for dual-target structure-based drug design and may not generalize well to other drug design tasks or domains.

High Computational Requirements

The computational requirements for training and deploying FuseDiff may be high, limiting its adoption in resource-constrained environments.

Expert Commentary

FuseDiff represents a significant advancement in the field of drug design, demonstrating the potential of diffusion models for jointly generating binding modes for multiple target pockets. While the model's performance is impressive, its limited generalizability and high computational requirements must be carefully considered in the context of real-world drug design applications. Furthermore, the development of polypharmacological therapies with improved efficacy and reduced resistance has significant implications for the pharmaceutical industry and healthcare systems. As such, FuseDiff is an important contribution to the field that warrants further investigation and exploration of its potential applications.

Recommendations

  • ✓ Future work should focus on adapting FuseDiff for other drug design tasks and domains to evaluate its generalizability.
  • ✓ Researchers should investigate the use of transfer learning and knowledge distillation to reduce the computational requirements of FuseDiff and make it more accessible to a wider range of users.

Sources

Related Articles