THFlow: A Temporally Hierarchical Flow Matching Framework for 3D Peptide Design
Dengdeng Huang, Shikui Tu
TL;DR
We address de novo 3D peptide design as a conditional generation problem p(Lig|Rec) across position, rotation, torsion, and amino-acid type. THFlow introduces a temporally hierarchical flow matching framework that uses a polynomial time-varying flow for the position modality while keeping orientation, torsion, and type flows linear and decoupled, enabling asynchronous convergence that mirrors actual docking dynamics. By incorporating interaction-related features and learning gradient fields with invariant point attention and transformers, THFlow achieves higher binding affinity and stability, and demonstrates faster, more accurate docking with reported gains such as a 7.13% affinity increase and roughly twofold more stable complexes. These results suggest THFlow provides a principled and efficient approach for sequence-structure co-design and re-docking, advancing peptide-based therapeutics by better capturing the temporal progression of peptide–protein interactions.
Abstract
Deep generative models provide a promising approach to de novo 3D peptide design. Most of them jointly model the distributions of peptide's position, orientation, and conformation, attempting to simultaneously converge to the target pocket. However, in the early stage of docking, optimizing conformation-only modalities such as rotation and torsion can be physically meaningless, as the peptide is initialized far from the protein pocket and no interaction field is present. We define this problem as the multimodal temporal inconsistency problem and claim it is a key factor contributing to low binding affinity in generated peptides. To address this challenge, we propose THFlow, a novel flow matching-based multimodal generative model that explicitly models the temporal hierarchy between peptide position and conformation. It employs a polynomial based conditional flow to accelerate positional convergence early on, and later aligns it with rotation and torsion for coordinated conformation refinement under the emerging interaction field. Additionally, we incorporate interaction-related features, such as polarity, to further enhance the model's understanding of peptide-protein binding. Extensive experiments demonstrate that THFlow outperforms existing methods in generating peptides with superior stability, affinity, and diversity, offering an effective and accurate solution for advancing peptide-based therapeutic development.