Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SculptDrug : A Spatial Condition-Aware Bayesian Flow Model for Structure-based Drug Design

Qingsong Zhong, Haomin Yu, Yan Lin, Wangmeng Shen, Long Zeng, Jilin Hu

TL;DR

SculptDrug tackles core challenges in structure-based drug design by integrating spatial conditioning with protein geometry. It leverages Bayesian Flow Networks with progressive denoising, augmented by a Boundary Awareness Block and a Hierarchical Encoder to enforce protein-surface constraints and multi-scale structural context. On the CrossDocked dataset, SculptDrug achieves superior binding affinity, drug-likeness, and structural plausibility, while reducing conformational strain and steric clashes relative to state-of-the-art baselines. This spatially aware, geometry-driven generative framework has potential to improve efficiency and reliability in ligand design for drug discovery.

Abstract

Structure-Based drug design (SBDD) has emerged as a popular approach in drug discovery, leveraging three-dimensional protein structures to generate drug ligands. However, existing generative models encounter several key challenges: (1) incorporating boundary condition constraints, (2) integrating hierarchical structural conditions, and (3) ensuring spatial modeling fidelity. To address these limitations, we propose SculptDrug, a spatial condition-aware generative model based on Bayesian flow networks (BFNs). First, SculptDrug follows a BFN-based framework and employs a progressive denoising strategy to ensure spatial modeling fidelity, iteratively refining atom positions while enhancing local interactions for precise spatial alignment. Second, we introduce a Boundary Awareness Block that incorporates protein surface constraints into the generative process to ensure that generated ligands are geometrically compatible with the target protein. Third, we design a Hierarchical Encoder that captures global structural context while preserving fine-grained molecular interactions, ensuring overall consistency and accurate ligand-protein conformations. We evaluate SculptDrug on the CrossDocked dataset, and experimental results demonstrate that SculptDrug outperforms state-of-the-art baselines, highlighting the effectiveness of spatial condition-aware modeling.

SculptDrug : A Spatial Condition-Aware Bayesian Flow Model for Structure-based Drug Design

TL;DR

SculptDrug tackles core challenges in structure-based drug design by integrating spatial conditioning with protein geometry. It leverages Bayesian Flow Networks with progressive denoising, augmented by a Boundary Awareness Block and a Hierarchical Encoder to enforce protein-surface constraints and multi-scale structural context. On the CrossDocked dataset, SculptDrug achieves superior binding affinity, drug-likeness, and structural plausibility, while reducing conformational strain and steric clashes relative to state-of-the-art baselines. This spatially aware, geometry-driven generative framework has potential to improve efficiency and reliability in ligand design for drug discovery.

Abstract

Structure-Based drug design (SBDD) has emerged as a popular approach in drug discovery, leveraging three-dimensional protein structures to generate drug ligands. However, existing generative models encounter several key challenges: (1) incorporating boundary condition constraints, (2) integrating hierarchical structural conditions, and (3) ensuring spatial modeling fidelity. To address these limitations, we propose SculptDrug, a spatial condition-aware generative model based on Bayesian flow networks (BFNs). First, SculptDrug follows a BFN-based framework and employs a progressive denoising strategy to ensure spatial modeling fidelity, iteratively refining atom positions while enhancing local interactions for precise spatial alignment. Second, we introduce a Boundary Awareness Block that incorporates protein surface constraints into the generative process to ensure that generated ligands are geometrically compatible with the target protein. Third, we design a Hierarchical Encoder that captures global structural context while preserving fine-grained molecular interactions, ensuring overall consistency and accurate ligand-protein conformations. We evaluate SculptDrug on the CrossDocked dataset, and experimental results demonstrate that SculptDrug outperforms state-of-the-art baselines, highlighting the effectiveness of spatial condition-aware modeling.

Paper Structure

This paper contains 21 sections, 15 equations, 7 figures, 3 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminaries
  4. Problem Definition
  5. Bayesian Flow Networks
  6. Spatial Condition-Aware Model
  7. Boundary Awareness Block
  8. Extracting the Protein Surface.
  9. Fitting the Key to the Lock.
  10. Hierarchical Encoder
  11. Generating Virtual Atoms for Protein Representation.
  12. Global Attention with Adaptive Edge Selection.
  13. Local Interaction Refinement with Distance-Aware Edges.
  14. Experiments
  15. Experimental Setup
  16. ...and 6 more sections

Figures (7)

  • Figure 1: Protein-ligand representation: (a) The protein ribbon model highlights the binding pocket, with close-ups showing the pocket in stick (top) and surface (bottom) representations, emphasizing spatial complementarity. (b) A symbolic "lock-and-key" analogy illustrates the specificity of protein-ligand binding.
  • Figure 2: (a) The generated ligand exhibits reasonable atomic distances when evaluated against protein atoms. (b) However, it erroneously penetrates the solvent-excluded surface, violating spatial plausibility.
  • Figure 3: Overview of the SculptDrug framework for ligand generation.
  • Figure 4: Impact of variants on SculptDrug 's performance.
  • Figure 5: Impact of the number of clusters $K$ on SculptDrug's performance.
  • ...and 2 more figures