Table 4 Comparison of TAPE benchmark results on three structure- or evolution-related tasks
From: Collectively encoding protein properties enriches protein language models
Model | Structure | Evolution | |
|---|---|---|---|
SS Prediction | Contact Prediction | Remote Homology | |
TAPE BERT | 0.73 | 0.36 | 0.21 |
Protein- BERT | 0.74 | – | 0.22 |
BERT medium | 0.74 | – | – |
String2Seq | – | – | 0.25 |
ProtBert | 0.80 | – | – |
MT-BERT | 0.75 | 0.39 | 0.32 |
MT-BCNN | 0.82 | 0.43 | 0.39 |
MT-BLSTM | 0.77 | 0.45 | 0.42 |