Growing Phylogenetic-like Trees
Preprocess:
- Describe compounds with keys
Process:
- Cluster active compounds with SOM
- Detect “hotspots” in the SOM
- Extract/learn filter-keys from hotspots
- Determine set of “useful” learned filter-keys
- Heuristic rules examples:
- Not a superset of another key in the same branch
- Not found in current or other branches of the tree
- Not a subset (improper) of its parent node’s key
Structural Family Identification