Summary
The distance between mean-pooled PLM embeddings of domains are negatively correlated with functional similarity of those domains (1).
See also
- Function prediction
- Protein language model embeddings are more predictive of homology than catalytic efficiency
1.
Ibtehaz N, Kagaya Y, Kihara D. Domain-PFP: Protein Function Prediction Using Function-Aware Domain Embedding Representations. openRxiv; 2023. Available from: https://doi.org/10.1101/2023.08.23.554486