Pauwels, E., Stoven,V., and Yamanishi, Y.,
Predicting drug side-effect profiles: a chemical fragment-based approach

Extracted drug chemical substructures and drug side-effects by Sparse CCA

• Extracted canonical components + top 30 scoring drugs.
  Sorted according to the canonical correlation
  

Predicted side-effects for unchatecterized drugs in DrugBank

• Summarized prediction result for uncharacterized drugs
  1st column: Drug ID
  2nd column: Pubchem compound ID
  3rd column: predicted side-effect
  4th column: prediction score
  

Result of 10-fold cross-validations

• ROC curve for evaluating global prediction accuracy
• Boxplot of the AUC scores for individual side-effects

Data sets

• list of all side-effect terms in this study
• drug representation matrix with PubChem chemical substructures
• drug representation matrix with SIDER side-effects
• drug representation matrix with PubChem chemical substructures for uncharacterized drugs (with no side-effect infromation)

Softwares: programs to carry out all analyses in the paper

• R functions and scripts
  The scripts are based on the "PMA" library in R.