Table 4 Average prediction accuracy of models trained with Proteinnet DSSP3 dataset
From: AttSec: protein secondary structure prediction by capturing local patterns from attention map
Model | TEST2018 | SPOT-2016 | SPOT-2016-HQ | SPOT-2018 | SPOT-2018-HQ |
|---|---|---|---|---|---|
PSIPRED-Single | 68.9 | 70.3 | 69.5 | 68.0 | 68.0 |
SPIDER3-Single | 72.6 | 72.0 | 72.2 | 71.3 | 70.8 |
ProteinUnet | 72.6 | – | – | – | – |
SPOT-1D-Single | 74.3 | 74.3 | 73.7 | 73.7 | 72.1 |
SPOT-1D (profile) | 86.2 | 81.7 | 83.1 | 80.4 | 82.0 |
AttSec(ours) | 86.6 | 81.6 | 81.9 | 81.5 | 81.3 |
SPOT-1D-Single (less than 1024) | 76.5 | 76.0 | 75.9 | 75.9 | 75.0 |
SPOT-1D-LMÂ (less than 1024) | 86.7 | 81.3 | 81.8 | 81.4 | 81.6 |
AttSec(ours)Â (less than 1024) | 86.5 | 81.7 | 81.9 | 81.6 | 81.4 |