Table 1.
Ten fold cross validated prediction accuracies on 2479 and 2640 sets for secondary structure prediction using SPINE X at three different steps
Dataset | 2479 | 2640 | |||
---|---|---|---|---|---|
Step | SS0 | SS1 | SS2 | SS2 | |
Assign. | SKSP+(DSSP) | SKSP+(DSSP) | SKSP+(DSSP) | (SKSP+) DSSP | DSSP |
Q3 | 81.5±0.4 (79.4) | 83.8±0.3 (81.0±0.4) | 83.8±0.4 (81.1) | (82.4) 82.1±0.4 | 82.0±0.5 |
QH | 86.6 (85.8) | 88.9 (87.8) | 88.9 (87.9) | (85.9) 86.4 | 86.6 |
QE | 74.0 (73.2) | 76.3 (74.6) | 76.4 (75.0) | (74.6) 75.6 | 75.3 |
QC | 80.9 (77.0) | 83.1 (78.0) | 83.0 (78.1) | (83.6) 81.9 | 81.5 |
The numbers in parentheses are the overall Q3 accuracy and accuracy for each secondary structure type according to the assignment method in parentheses but where the weights were trained by using the assignment not in parentheses. Error bars give the standard deviation over the ten folds.