Table 3.
Graph-based models for BAP.
| Models | Ligand Input Representation | Protein Input Representation | Ligand Feature Learning | Protein Feature Learning |
|---|---|---|---|---|
| GraphDTA [4] | Molecular graph | one-hot encoding of AA seq | GCN\GAT\GIN\GATGCN | CNN |
| DeepGLSTM [9], | Molecular graph | Lebel encoding of AA seq | Multiblock GCN | Bi-LSTM |
| GDGRU-DTA [38], | Molecular graph | Lebel encoding of AA seq | GNN | GRU/BiGRU |
| EmbedDTI [60] | Atom graph and substructure graph | AA seq | GCN with attention | 1D CNN |
| DeepGS [59] | Molecular graph + Smi2Vec | Prot2Vec | GAT + BiGRU | CNN |
| Dgraph-DTA [61] | Molecular graph | Protein graph | GCN | GCN |
| X-DPI [62] | Molecular graph + Mol2vec embedding | Protein graph + TAPE embedding | GCN | GCN |
| WGNN-DTA [63] | Molecular graph | Weighted protein graph | GCN/GAT | GCN/GAT |
| DGDTA [36] | Molecular graph | Lebel encoding of AA seq | Dynamic GAT + GCN | Bi-LSTM + CNN |
| PSG-BAR [64] | Molecular graph | Protein graph | Residual GAT | Residual GAT |
| GraphBAR [66] | Binding complexes graph | Binding complexes graph | Multiblock GCN | Multiblock GCN |
| APMNet [67] | Binding complexes graph | Binding complexes graph | GCN | GCN |
| LGN [71] | Molecular graph + IFP | Complexes graph | GIN | GNN |
| PLANET [72] | Molecular graph | Protein graph | GNN | EGCL |
| GraphscoreDTA [73] | Molecular graph | Protein graph + Interaction Graph | GNN-GRU | GNN + GNN-GRU |
| GraphATT-DTA [74] | Molecular graph | Lebel encoding of AA seq | GNN (GAT/GIN/GCN/MPNN/DMPNN) | 1D CNN |
| AttentionMGT-DTA [75] | Molecular graph | Protein graph | Graph transformer | Graph transformer |
| GLCN-DTA [76] | Molecular graph | Protein graph | GLCN | GLCN |
| IEDGEDTA [77] | Molecular graph | Protein graph | Edge-GCN | 1D-GCN |
| SAG-DTA [78] | Molecular graph | Lebel encoding of AA seq | GCN | 1D CNN |
Abbreviations: AA seq: Amino acid sequence; GNN — graph neural network; MLP — multilayer perceptron; GCN: graph convolutional network; GAT: graph attention network; GIN: graph isomorphism network; CNN: convolutional neural network; FC: fully connected layer; GRU: gated recurrent unit; EGCL: Equivariant graph convolutional layer, MPNN: message passing neural network, IFP: interaction fingerprint; GLCN: Graph learning convolutional network.