Table 1 Specifications of the studied tools; KDeep, KDeep + , DeepRKE-, DeepRKE + , GraphProt, DeepBind, iDeepS, DanQ, iDeepV, deepnet-rbp, mmCNN
Methods | Strand type | Input type | Encoding method | Predictor architecture |
|---|---|---|---|---|
KDeep | RNA/DNA | Primary | 2Lk | CNN + LSTM |
KDeep +  | DNA | Primary | 2Lk | CNN + LSTM + Attention |
DeepRKE- | RNA | Primary | Word2vec | CNN + LSTM |
DeepRKE +  | RNA | Primary + secondary | Word2vec | CNN + LSTM |
GraphProt | RNA | Primary + secondary | Graph | SVM |
DeepBind | DNA/RNA | Primary | One-hot | CNN |
iDeepS | RNA | Primary + secondary | One-hot | CNN + LSTM |
DanQ | DNA | Primary | One-hot | CNN + LSTM |
iDeepV | RNA | Primary | Word2vec | CNN |
deepnet-rbp | RNA | Primary + secondary + tertiary | (Primary & Secondary) encode by replicated software model RNA, tertiary structure encoded by RNA 3D Motif | CNN |
mmCNN | RNA | Primary + secondary + tertiary | Primary (one-hot) Secondary (Structure Probability Matrix) | CNN |