Skip to main content

Table 5 Results of DistDom, DeepDom, and FUPred on the multi-domain targets in the CASP14 dataset in terms of average per target F1, specificity, sensitivity and Matthew’s Correlation Coefficient with their confidence intervals at 95% confidence

From: Multi-head attention-based U-Nets for predicting protein domain boundaries using 1D sequence features and 2D distance maps

Method

F1 score

Specificity

Sensitivity

MCC

DistDom

0.263 (0.147,0.378)

0.936 (0.913,0.959)

0.403 (0.220,0.586)

   0.249 (0.123,0.376)

DeepDom

0.134 (0.040,0.227)

0.853 (0.792,0.913)

0.123 (0.031,0.215)

− 0.033 (− 0.14,0.07)

FUPred

0.203 (0.080,0.318)

0.960 (0.941,0.978)

0.266 (0.122,0.410)

    0.186 (0.070,0.301)