Table 1 Two-branch methods
From: A comparison of embedding aggregation strategies in drug–target interaction prediction
Method | Compound representation | Compound branch | Protein representation | Protein branch |
|---|---|---|---|---|
DeepConv-DTI [4] | Morgan | MLP | Amino acid sequence | CNN |
DeepDTA [5] | SMILES | CNN | Amino acid sequence | CNN |
Shin et al. [6] | SMILES | Transformer | Fasta | CNN |
MDeePred [7] | ECFP4 | MLP | Multi-channel protein features | CNN |
Chen et al. [8] | Molecular graph | GNN | Amino acid sequence | BERT |
Torng et al. [9] | Molecular graph | GCNN | Protein graph | GCNN |
SSnet [10] | Fingerprint | MLP | Backbone curvature, torsion | CNN |
Tsubaki et al. [11] | Molecular graph | GNN | Amino acid sequence | CNN |
Kang et al. [12] | SMILES | BERT | Amino acid sequence | BERT |
Nguyen et al. [13] | Molecular graph | GCN, GAT, GIN, GAT-GCN | Amino acid sequence | CNN |
Mingjian et al. [14] | Molecular graph | GNN | Protein graph | GNN |
DeepPurpose [15] | Various encodings | MLP, CNN, CNN–RNN... | Various encodings | MLP, CNN, CNN–RNN... |