Table 1 Prediction methods
From: DGDTA: dynamic graph attention network for predicting drug–target binding affinity
Method | Published time | Model | Summary |
|---|---|---|---|
SimBoost [11] | 2016 | Gradient boosting regression trees | Predicting continuous values of binding affinities of compounds and proteins |
KronRLS [16] | 2018 | Multiple kernel learning | The first method for time- and memory-efficient learning with multiple pairwise kernels |
DeepDTA [8] | 2018 | CNN | Processing protein sequences and compound 1D representations using convolutional neural networks |
PADME [13] | 2018 | DNN | The first to combine Molecular Graph Convolution for compound featurization with protein descriptors |
WideDTA [17] | 2019 | CNN | Combining four different textual pieces of information related to proteins and ligands |
MT-DTI [18] | 2019 | Transformers + CNN | Proposing a new molecule representation based on the self-attention mechanism |
GANsDTA [24] | 2019 | GAN + CNN | Effectively learning valuable features from labeled and unlabeled data |
DeepGS [19] | 2020 | GAT + Bi-GRU | Extracting the topological information of the molecular map and the local chemical context of the drug |
rzMLP [22] | 2021 | gMLP + ReZero | Use MHM block for multiple protein and ligand representations and rzMLP block to aggregate concatenated protein-ligand pair representations |
EnsembelDLM [23] | 2021 | Multiple deep networks | Aggregating predictions from multiple deep neural networks |
GraphDTA [26] | 2021 | GIN + CNN | Introducing multiple models of graph neural networks |