Demystifying Multimodal Biomolecular Co-design With Intrinsic Geodesic Coupling
2026-06-01 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors study how to better design biomolecules like proteins and small molecules by looking at their sequence and 3D structure together. They point out that previous models treat these two parts at the same time in a fixed way, which can cause problems during training. To fix this, the authors propose GeoCoupling, a method that learns the best timing to connect these parts while designing molecules. Their approach leads to biomolecules that are more physically accurate and varied compared to older methods.
biomoleculesprotein designsmall-molecule ligandsgenerative models3D structuresequence-structure relationshiptemporal couplingdrug designmodel training
Authors
Keyue Qiu, Xintong Wang, Zhilong Zhang, Hao Zhou, Wei-Ying Ma
Abstract
Biomolecules such as proteins and small-molecule ligands play a central role in biological systems, arising from the tight interplay between sequence and three-dimensional structure. Recent generative models for biomolecular co-design aim to capture this interplay by jointly modeling coupled modalities. However, existing approaches largely adopt a parallel execution of marginal generative processes, implicitly enforcing fixed synchronous coupling. We argue that a critical but overlooked degree of freedom lies in how these marginal processes are temporally coupled during training and generation, where inappropriate coupling can introduce high-variance supervision and inconsistent intermediate states, affecting modality consistency. To address this, we introduce GeoCoupling, a systematic framework that optimizes for temporal couplings between heterogeneous modalities. Empirical results across structure-based drug design and unconditional protein design demonstrate the learned couplings consistently outperform synchronous and randomly coupled baselines, yielding biomolecules with improved physical validity and diversity.