Performance Measures
Analyse datasets extracted from WDI
- 7 activity classes mixed with 3500 compounds selected at random
Generate reduced graphs at different levels
Convert reduced graphs to "pseudo-smiles" using different heavy atoms to distinguish node types
Clustering
- Check for identical reduced graphs by comparing pseudo-smiles
NN lists
- Generate Daylight fingerprints from the pseudo-smiles
- Calculate NN list of target structure
- Count number of actives occurring in top 300 positions of ranked list (A@300)