DataSAIL Addendum Splits

Here we store and distribute the splits computed for the Addendum to our DataSAIL manuscript. This contains leakage-minimized DataSAIL splits for 14 of the 17 datasets in the MoleculeNet molecular property prediction benchmark [DOI: 10.48550/arXiv.1703.00564], the Leak-Proof PDBBind database [DOI: 10.48550/arXiv.2308.09639], the Protein Ligand INteraction Dataset and Evaluation Resource (PLINDER) [DOI: 10.1101/2024.07.17.603955], the Protein INteraction Dataset and Evaluation Resource (PINDER) [DOI: 10.1101/2024.07.17.603980], and the "gold standard" dataset of human protein-protein interactions by Bernett et al. [DOI: 10.1093/bib/bbae076], All splits were calculated using DataSAIL v1.2.1. The uploaded files contain both the DataSAIL splits and the dataset's provided splits.