Ryusuke Sawada, Yuko Sakajiri, Tomokazu Shibata, and Yoshihiro Yamanishi,
Predicting therapeutic and side effects from drug binding affinities to human proteome structures

Supplemental result

Supplemental data and profiles in the manuscript

• Docking site information
  A list of docking site detected by AutoSite including size and location information. These are used for docking simulation with the AutoDock Vina.
  
• Proteome-wide binding affinity score profiles
  Proteome-wide binding affinity score (PBAS) profiles were developed from results of large-scale docking simulations for 7,582 drugs with 19,135 protein structures. Profile scores represent the protein-compound binding free energy that was calculated using AutoDock Vina. Protein structures and drugs were indicated as AlphaFold Protein Structure Database ID and KEGG DRUG ID, respectively.
  
• Profiles for side-effect predictive models
  Four profiles were used as feature vectors for the drugs to construct the side-effect predictive models for comparing the side-effect prediction performance. The archive file includes following files:
  i) fingerprint.tsv
  Fingerprint profiles representing the substructures of drugs. Profile consists of 429 drugs and Morgan fingerprints 2,048 features.
  ii) tess.tsv
  Target protein profiles for drugs using the prediction method of target estimation with similarity search (TESS). Profile consists of 429 drugs and 4643 target protein features. Profile scores represent the drug-protein interaction score using TESS, which estimate the interaction based on the chemical structural similarities to the compound-protein interaction dataset.
  iii) telr.tsv
  Target protein profiles for drugs using the prediction method of target estimation with logistic regression (TELR). Profile consists of 429 drugs and 4643 target protein features. Profile scores represent the drug-protein interaction score using TELR, which estimate the interaction using the machine learning classification model trained by compound-protein interaction dataset.
  iv) pbas.af.tsv
  PBAS profiles consists of 429 drugs with 19,135 protein structures.