Table 1.

Representative allosteric site prediction methods

Features	Methods	Datasetsa	Refs
Static pocket features	Naïve Bayes and neural networks	ASD and ASBench	[70]
Static pocket features	GCNN with XGBoost	ASD	[71]
Static pocket features	Automated machine learning	ASD and ASBench	[72]
Static pocket features	Random forest	ASD b	[73]
Static pocket features	Support vector machine	ASD	[74]
Pocket features with NMA perturbation	Support vector machine	ASBench	[75]
Pocket features with NMA perturbation	Logistic regression	ASBench	[76]
Features at residue level	Random forest	ASBench	[77]
Crystal additive location	DBSCAN	ASD and ASBenchc	[78]

^aThe original datasets used to obtain allosteric site data. The data was filtered for high-resolution and non-redundant structures, individually.

^bThe PDBbind database was used to obtain information on orthosteric sites.

^cThe RCSB protein data bank was used to obtain protein-crystallographic additives complexes.