HotPatch's statistical parameters for kinases are derived by training on a set of structures in this class.

In training on their known functional sites, the sites were defined by:

• Annotations as functional residues (involved in catalysis, substrate recognition, binding etc.) in the SFR Database, which were mostly obtained from the literarture.
• Proximity of 4Å to a substrate (if present in the structure), or to a known surrogate of the substrate (e.g. inhibitor known to bind where substrate normally does).

Size of training set: 25 structures (23 with hydrogens added, for electrostatics & neural networks).