Piloting Structure-Based Drug Design via Modality-Specific Optimal Schedule

Keyue Qiu, Yuxuan Song, Zhehuan Fan, Peidong Liu, Zhe Zhang, Mingyue Zheng, Hao Zhou, Wei-Ying Ma

TL;DR: Twisted probability path induced by noise schedule with the optimal VLB

Abstract: Structure-Based Drug Design (SBDD) is crucial for identifying bioactive molecules. Recent deep generative models are faced with challenges in geometric structure modeling. A major bottleneck lies in the twisted probability path of multi-modalities—continuous 3D positions and discrete 2D topologies—which jointly determine molecular geometries. By establishing the fact that noise schedules decide the Variational Lower Bound (VLB) for the twisted probability path, we propose VLB-Optimal Scheduling (VOS) strategy in this under-explored area, which optimizes VLB as a path integral for SBDD. Our model effectively enhances molecular geometries and interaction modeling, achieving state-of-the-art PoseBusters passing rate of 95.9\% on CrossDock, more than 10\% improvement upon strong baselines, while maintaining high affinities and robust intramolecular validity evaluated on held-out test set. Code is available at https://github.com/AlgoMole/MolCRAFT.

Lay Summary:

**Problem**: Developing new drugs requires creating molecules that can precisely fit into target proteins, similar to finding the right key for a lock. Current methods struggle to design these molecules accurately because they must handle two different aspects simultaneously like solving a 2D/3D jigsaw puzzle: the 3D arrangement of atoms *(the shape of the key)* and the 2D connections between different types of atoms *(the material of the key)*. This often results in unrealistic molecular designs that either don’t fit well or are chemically unstable.

**Solution**: We developed VOS, a strategy that acts like a “smart roadmap” during the generative process, ensuring both the 3D geometry and 2D structure evolve harmoniously in an analytical way. Think of it as teaching the generative model to refine how it builds molecules, not just what it builds.

**Impact**: Integrated with advanced framework, our MolPilot significantly improves the physical plausibility of AI-generated drug candidates, with 95.9% PoseBusters passing rate (10%+ higher than previous SOTA). These molecules also maintain strong binding to target proteins, potentially accelerating drug discovery by providing scientists with more reliable starting points for development.

Link To Code: https://github.com/AlgoMole/MolCRAFT

Primary Area: Applications->Chemistry, Physics, and Earth Sciences

Keywords: Structure-Based Drug Design, Bayesian Flow Network, Noise Schedule

Flagged For Ethics Review: true

Submission Number: 1546