Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper investigates the impact of compositional sequence embeddings on protein-protein interaction networks by training Event2Vec, an additive sequence embedding model, on random walks from the human STRING interactome. They demonstrate that Event2Vec, which enforces compositional structure, significantly outperforms a DeepWalk baseline in pathway coherence (30.2x vs 2.9x above random) and functional analogy accuracy (0.966 vs 0.650). The results suggest that enforcing compositionality in embeddings enhances relational and compositional reasoning within biological networks.
Enforcing compositional structure in protein network embeddings yields a 10x boost in functional analogy accuracy, suggesting a new path to understanding biological function.
In this work, we study whether enforcing strict compositional structure in sequence embeddings yields meaningful geometric organization when applied to protein-protein interaction networks. Using Event2Vec, an additive sequence embedding model, we train 64-dimensional representations on random walks from the human STRING interactome, and compare against a DeepWalk baseline based on Word2Vec, trained on the same walks. We find that compositional structure substantially improves pathway coherence (30.2$\times$ vs 2.9$\times$ above random), functional analogy accuracy (mean similarity 0.966 vs 0.650), and hierarchical pathway organization, while geometric properties such as norm--degree anticorrelation are shared with or exceeded by the non-compositional baseline. These results indicate that enforced compositionality specifically benefits relational and compositional reasoning tasks in biological networks.