BIPSPI+
Data sets
Tab separated files that contains the following columns: PDB_id, chainId(s)_partner1, chainId(s)_partner2, scopes_partner1, scopes_partner2. If PDB_id is written in capital letters, then it belongs to the testing set (either PP Docking Benchmark v5 or HOe dataset). If is written in lowercase, then it belongs to the training set. Independently to which set they belog, during the 10-fold cross validation, independence must be ensured by the user checking that any of the compounds in the training contain any of the same scope family pairs than the compounds in the testing split.
- Test and train dataset for homo-complexes (HOe+HODt)
- Test and train dataset for hetero-dimers(Bv5+HEDt)
- Test and train dataset for hetero-multimers (Bv5+HEMt)
- Test and train dataset for hetero and homo dimers (Bv5+HOe+HODHEDt)
Results
Cross-validation predictions for test complexes from model trained using different datasets
- Trained on homo-complexes tested in homodimers (HOe+HODt)
- Trained on heterodimers tested in hetero-complexes(Bv5+HEDt)
- Trained on hetero-complexes tested in hetero-complexes(Bv5+HEMt)
- Trained in hetero and homo-complexes and tested in hetero-complexes and homodimers (HOe+Bv5+HEHODt)
- Trained on heterodimers tested in homodimers (HOe+HEDt)
- Trained on hetero-complexes tested in homodimers (HOe+HEMt)
BIPSPI v1
- Predictions for Benchmark v3 leave-one-complex-out using sequential and structural features.
- Predictions for Benchmark v3 leave-one-complex-out using sequential features.
- Predictions for Benchmark v5 leave-one-complex-out using sequential and structural features.
- Predictions for Benchmark v5 leave-one-complex-out using sequential features.
- PDB files compiled in dimers benchmark
- Predictions for dimers benchmark leave-one-complex-out using sequential and structural features.
- Predictions for dimers benchmark leave-one-complex-out using sequential features.
- Capri targets evaluated in PAIRPred
- Predictions for capri targets employed in PAIRPred using sequential and structural features. Results for ISPRED4 comparison:
- Capri targets evaluated in ISPRED4
- Predictions for capri targets employed in ISPRED4 using sequential and structural features.
- Predictions using Benchmark v 4 complexes as train and Benchmark v5 complexes not included in Benchmark v 4 as test
Benchmark v3
Publication: Hwang, H. et al. (2008) Protein-Protein Docking Benchmark Version 3.0, Proteins, 73(3), 705-709.Results for PAIRPred comparison:
Benchmark v5
Benchmark websitePublication: Vreven, T. et al. (2015) Updates to the Integrated Protein-Protein Interaction Benchmarks: Docking Benchmark Version 5 and Affinity Benchmark Version 2, Journal of Molecular Biology, 427(19), 3031-3041.