Supplements
Yamanishi,Y., Kotera,M., Kanehisa,M., and Goto,S.
Drug-target interaction prediction from chemical, genomic, and pharmacological data in an integrated framework
Predicted drug-target interaction networks
- All newly predicted compound-protein interacion pairs with annotation information (text file)
Note that the top scoring predicted drug-target interaction pairs with the upper one percent score percentile are shown.
The 1st column and the 2nd column indicate the protein and the compound in the predicted interaction pair, respectively.
The 3rd column indicates the candidate score in the prediction.
The 4th column and the 5th column indicate the protein and the compound in the mimicked interaction pairs.
The mimicked interaction pair means a known drug-target interaction pair (in the gold standard data) which is the closest to the predicted drug-target interaction pair in the feature space.
Datasets used in this study
- Adjacency matrix of the gold standard drug-target interaction data (text file)
Note that this is an adjacency matrix representation of drug-target interactions obtained from KEGG BRITE, SuperTarget, DrugBanck, and BRENDA.
- Chemical structure similarity matrix for compounds in gold standard data (text file)
Note that the similarity scores are computed by the SIMCOMP score (Hattori et al, J.Ame.Chem.Soc, 2003)
- Genomic sequence similarity matrix for proteins in gold standard data (text file)
Note that the similarity scores are computed by the normalized SmithWaterman score (Smith and Waterman, J.Mol.Bio, 1981)
- Pharmacological similarity matrix for compounds in gold standard data (text file)
Note that the similarity scores are based on the pharmaceutical effect category in the JAPIC database.
- Chemical structure similarity matrix: full (test+training) compounds vs training compounds (text file)
Note that the similarity scores are computed by the SIMCOMP score (Hattori et al, J.Ame.Chem.Soc, 2003)
- Genomic sequence similarity matrix: full (test+training) compounds vs training proteins (text file)
Note that the similarity scores are computed by the normalized SmithWaterman score (Smith and Waterman, J.Mol.Bio, 1981)
- Pharmacological similarity matrix: all possible compounds vs themselves (text file)
Note that the similarity scores are based on the weighted cosine correlation coefficient.
- Original pharmacological profiles (text file)
Note that pharmacological keywrods for each drug are obtained from the JAPIC database. (May 2009)