Enhancing Protein-Protein Interaction Prediction with Hierarchical Motif-based Multimodal Protein Embedding
Zaifei YANG, Samuel Ping-Man Choi, James Kwok · 2026
Read Enhancing Protein-Protein Interaction Prediction with Hierarchical Motif-based Multimodal Protein Embedding by Zaifei YANG, Samuel Ping-Man Choi, James Kwok (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-02629.txt.
Description
Yang and co-authors build a protein encoder that respects how proteins are organized. Protein-protein interactions drive many biological processes. Past predictors miss two things. They ignore meso-scale motifs, the mid-level parts that regulate interactions. They also fuse sequence, structure, and function poorly. Their encoder, MMM-PPI, works bottom-up across three scales.
At the micro scale, it encodes three kinds of residue features. At the meso scale, a new motif encoder groups residues into spatially informed motif embeddings. At the macro scale, motifs merge into protein embeddings. Motif importance and inter-modal correlations are modeled jointly. The pretrained encoder works off the shelf for large-scale prediction. Tests on several datasets beat top multi-label PPI models. The lead holds under hard data partitions and scarce data. Code is released in public.
Catalog record
| CREATOR | Zaifei YANG, Samuel Ping-Man Choi, James Kwok |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 73 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.02629