PocketVE: Stable and Property-Guided Structure-Based Drug Design with Variance-Exploding Diffusion
Abstract
Protein-conditioned 3D molecule generation is a central challenge in structure-based drug design, requiring a balance between pocket compatibility, molecular properties, and physical geometry. We propose \textbf{PocketVE}, a protein-pocket-conditioned variance-exploding (VE) diffusion framework that couples stable coordinate denoising with inference-time property guidance. Specifically, PocketVE combines an EDM-style training and sampling setup for 3D denoising, classifier-free guidance for multi-property steering, and adaptive protein perturbation as a training-time pocket regularizer.On the CrossDocked2020 benchmark under GenBench3D, PocketVE improves Valid3D from 58.6 to 80.6 and reduces strain energy from 457.4 to 127.9 relative to its guided TAGMol architectural parent; relative to TargetDiff, it attains comparable Valid3D with lower strain energy (127.9 vs.\ 306.0), while retaining competitive docking and molecular-property scores under moderate guidance. A guidance-scale study shows that moderate guidance gives a favorable balance between target-related objectives and geometric quality, whereas stronger guidance can degrade geometry and distributional fidelity. Pocket-permutation and PoseCheck diagnostics further support pocket-specific spatial compatibility with reduced steric conflicts. Overall, the results suggest that geometric stability and inference-time property guidance should be considered as coupled design objectives.
Related
- Structure–Activity–Property Relationships of MZ1-Based Trivalent PROTACs Incorporating a 1,2,5-Trisubstituted Benzene Linker
- Biosilica from Agricultural Residues for Hazardous Contaminant Removal and Soil Remediation: Synthesis Routes, Structure–Property Relationships, and Environmental Relevance
- Rare-Earth Orthoferrites: Synthesis–Structure–Property Relationships, Functional Applications, and Future Perspectives
- Analysis of the Effects of Silica Surface Structure on the Local Diffusion and Hydrogen Bonding of Water via Molecular Dynamics.
- Mathematical Invariant-Enabled Topological Neural Networks for Molecular and Materials Property Prediction
BibTeX
@misc{zhang2026pocketve,
title = {{PocketVE: Stable and Property-Guided Structure-Based Drug Design with Variance-Exploding Diffusion}},
author = {Peining Zhang and Jinbo Bi},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.08101},
eprint = {2609.08101},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.08101},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - PocketVE: Stable and Property-Guided Structure-Based Drug Design with Variance-Exploding Diffusion
AU - Peining Zhang
AU - Jinbo Bi
PY - 2026
DA - 2026/09/27
PB - arXiv
DO - 10.48550/arxiv.2609.08101
UR - https://arxiv.org/abs/2609.08101
AB - Protein-conditioned 3D molecule generation is a central challenge in structure-based drug design, requiring a balance between pocket compatibility, molecular properties, and physical geometry. We propose \textbf{PocketVE}, a protein-pocket-conditioned variance-exploding (VE) diffusion framework that couples stable coordinate denoising with inference-time property guidance. Specifically, PocketVE combines an EDM-style training and sampling setup for 3D denoising, classifier-free guidance for multi-property steering, and adaptive protein perturbation as a training-time pocket regularizer.On the CrossDocked2020 benchmark under GenBench3D, PocketVE improves Valid_3D from 58.6 to 80.6 and reduces strain energy from 457.4 to 127.9 relative to its guided TAGMol architectural parent; relative to TargetDiff, it attains comparable Valid_3D with lower strain energy (127.9 vs.\ 306.0), while retaining competitive docking and molecular-property scores under moderate guidance. A guidance-scale study shows that moderate guidance gives a favorable balance between target-related objectives and geometric quality, whereas stronger guidance can degrade geometry and distributional fidelity. Pocket-permutation and PoseCheck diagnostics further support pocket-specific spatial compatibility with reduced steric conflicts. Overall, the results suggest that geometric stability and inference-time property guidance should be considered as coupled design objectives.
ER -