Reproduce “Learning the Interaction Prior for Protein-Protein Interaction Prediction: A Model-Agnostic Approach” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ziqi Gao, Chenyi Zi, Zijing Liu, Ziqiao Meng, Yu Li, Jia Li
The L3-PPI framework introduces a model-agnostic classification head that enhances protein-protein interaction prediction by explicitly incorporating biological complementarity priors through learnable length-3 path patterns. It consistently improves prediction accuracy for existing deep learning methods, especially in generalization-focused scenarios, by reformulating pairwise prediction into a graph-level classification task.
Key insights
- The "L3 rule," which states that the number of length-3 paths between proteins correlates with interaction likelihood, is empirically validated as a strong indicator across PPI datasets, often outperforming other path lengths.
- Incorporating explicit biological complementarity priors into deep learning models via a structured, graph-level prompt tuning approach significantly improves PPI prediction, particularly in challenging generalization settings.
- Reformulating pairwise PPI prediction into a graph-level classification task over dynamically generated "pattern graphs" allows for effective transfer and application of pre-trained biological knowledge (L3 pattern validity).
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.