Table 2.

List of deep learning methods to predict protein–ligand binding sites

SN	Approach	Techniques involved	Feature	Database used	Year
1	DeepCSeqSite [95]	Deep convolutional neural network	Proposed sequence-based approach for ab initio protein–ligand binding residue prediction.	BioLip	2019
2	DELIA [96]	Hybrid Deep neural network + bidirectional long short-term memory network	Designed hybrid deep neural network is to integrate 1D sequence-based features with 2D structure-based amino acid distance matrices.	BioLip + ATPBind	2020
3	Kalasanty [97]	3D convolutional neural network	Designed model based on U-Net’s architecture.	sc-PDB [98]	2020
4	DeepSurf [99]	Deep convolutional neural network + ResNet	Proposed surface-based deep learning approach for protein–ligand binding residue prediction.	scPDB	2021
5	PUResNet [100]	ResNet	Based on deep ResNet + novel data cleaning process.	scPDB	2021