Table 1.

Common structure-based methods to predict ligand binding pocket in proteins.

Category	Search Approach	Methods
Geometric	Grid	CAVER [51], CAVIAR [49], DoGSite [14], ghecom [57], KVFinder [48], LIGSITE [12], LIGSITEcsc [53], McVol [58], POCKET [13], PocketDepth [55], PocketPicker [54] VICE [56], VOIDOO [52], VOLSITE [47]
	Alpha-shape	APROPOS [59], CAST [6], CASTp [60], Fpocket [16],
	Spherical probes	DEPTH [61], HOLE [62], HOLLOW [63], PHECOM [57], PASS [64], Roll [65], SURFNET [66], SURFNET-ConSurf [67], Xie and Bourne [33]
	Other	MSPocket [68], SplitPocket [69]
Energetic	Grid	AutoLigand [70], DrugSite [71], FTSite [72], PocketFinder [73], Q-SiteFinder [74], SITEHOUND [75], SiteMap [76], pocket-finder [77], GRID [78]
	Spherical probes	dPredGB [79], Morita et al. [80]
	Other	Gaussian Network Model [81]
Data-driven	Machine learning	GRaSP [82], MCSVMBs [83], P2Rank [15], PRANK [84], SCREEN [85]
	Deep learning	PoinSite [86], DeepPocket [87], PUResNet [88], DeepSurf [89], BiteNet [90], Jiang et al. [91], DeepSite [92], ISMBLab-LIG [91]