metadata
license: mit
This is a dataset of 10,000 interacting pairs of proteins obtained from UniProt, and clustered using methods explain in this blog post. Note, cluster 0 is over-represented in this dataset and this should be considered when creating train/test splits with this data.