Supplements
Takarabe,M., Kotera,M., Nishimura,Y., Goto,S. and Yamanishi,Y.
Drug target prediction using adverse event reporting systems: a pharmacogenomic approach
Supplemental Table 1: Results of AUPR (Area under Precision-Recall curve) scores
- Summary of AUPRs in the 3-fold cross-validation (text file).
Common: common durgs associated interaction data.
All: all drugs associated interaction data.
Threshold: chemical similarity threshold in the clustering to select representative drugs.
Supplemental Table 2: Performance comparison of different algorithms
- Summary of AUCs and computational cost in the 3-fold cross-validation (text file).
Newly predicted drug-target interactions
- Predicted drug-target pairs (text file), where the top 3000 scoring pairs are shown.
Drug (1st column): drug in the predicted interaction pair, respectively.
Target (2nd column): target protein in the predicted interaction pair.
Score (3rd column): prediction score.
MimicD (4th column): training drug (in the training data) which is the colosest to the drug in the predicted pair.
MimicT (5th column): training target (in the training data) which is the colosest to the target in the predicted pair.
MiDsim (6th column): similarity between Drug and MimicD.
MiTsim (7th column): similarity between Target and MimicT.
Type (8th column): prediction type, where t_t: training drug vs training target, p_t: new drug vs training target, t_p: training drug vs new target, and p_p: new drug vs new target.
- Comparison of predicted drug-target pairs between AERS and the others (SIDER, JAPIC, CHEM), where the top 10000 scoring pairs are shown.
Note that the following 3 columns are added to the result of AERS-based predictions.
R-SIDER: rank in the SIDER-based predictions.
R-JAPIC: rank in the JAPIC-based predictions.
R-CHEM: rank in the CHEM-based predictions.
"NA" means that the pair cannot be predicted by each of the other methods.
Newly predicted drug-target interactions with annotation information
- Predicted drug-target pairs (text file)
Drug (1st column): drug in the predicted interaction pair, respectively.
Target (2nd column): target protein in the predicted interaction pair.
Score (3rd column): prediction score.
MimicD (4th column): training drug (in the training data) which is the colosest to the drug in the predicted pair.
MimicT (5th column): training target (in the training data) which is the colosest to the target in the predicted pair.
MiDsim (6th column): similarity between Drug and MimicD.
MiTsim (7th column): similarity between Target and MimicT.
Type (8th column): prediction type, where t_t: training drug vs training target, p_t: new drug vs training target, t_p: training drug vs new target, and p_p: new drug vs new target.
- Comparison of predicted drug-target pairs between AERS and the others (SIDER, JAPIC, CHEM) with annotation information (text file)
Note that the following 3 columns are added to the result of AERS-based predictions.
R-SIDER: rank in the SIDER-based predictions.
R-JAPIC: rank in the JAPIC-based predictions.
R-CHEM: rank in the CHEM-based predictions.
"NA" means that the pair cannot be predicted by each of the other methods.
Common drugs-associated data
- Adjacency matrix of drug-target interactions for common drugs (text file)
Note that the information is obtained from KEGG DRUG.
- AERS-freq-based pharmacological similarity matrix for drugs (text file)
Note that the similarity scores are based on side-effect terms in AERS.
- AERS-bit-based pharmacological similarity matrix for drugs (text file)
Note that the similarity scores are based on side-effect terms in AERS.
- SIDER-based pharmacological similarity matrix for drugs (text file)
Note that the similarity scores are based on side-effect terms in SIDER.
- JAPIC-based pharmacological similarity matrix for drugs (text file)
Note that the similarity scores are based on side-effect terms in JAPIC.
- Chemical structure similarity matrix for drugs (text file)
Note that the similarity scores are computed by the graph kernel (Mahe et al, J.Chem.Inf.Model, 2005)
- Genomic sequence similarity matrix for target proteins (text file)
Note that the similarity scores are computed by the local alignment kernel (Saigo et al, Bioinformatics, 2004)
All drugs-associated data
- Adjacency matrix of drug-target interactions for all drugs with target information (text file)
Note that the information is obtained from KEGG DRUG.
- AERS-freq-based pharmacological similarity matrix for drugs (text file)
Note that the similarity scores are based on side-effect terms in AERS.
- AERS-bit-based pharmacological similarity matrix for drugs (text file)
Note that the similarity scores are based on side-effect terms in AERS.
- SIDER-based pharmacological similarity matrix for drugs (text file)
Note that the similarity scores are based on side-effect terms in SIDER.
- JAPIC-based pharmacological similarity matrix for drugs (text file)
Note that the similarity scores are based on side-effect terms in JAPIC.
- Chemical structure similarity matrix for drugs (text file)
Note that the similarity scores are computed by the graph kernel (Mahe et al, J.Chem.Inf.Model, 2005)
- Genomic sequence similarity matrix for target proteins (text file)
Note that the similarity scores are computed by the local alignment kernel (Saigo et al, Bioinformatics, 2004)