Skip to main content

Table 1 Performance on HARD CASP10 targets

From: A study and benchmark of DNcon: a method for protein residue-residue contact prediction using deep networks

 

Acc. Top L/10 (SE)

Acc. Top L/5 (SE)

Acc. Top L (SE)

Method (GroupID)

Long

Medium

Long

Medium

Long

Medium

IBGteam [DL] (305)

0.263 (0.066)

0.356(0.085)

0.208 (0.050)

0.292(0.063)

0.117 (0.024)

0.180(0.027)

DNcon (222)

0.244 (0.039)

0.442(0.072)

0.207 (0.029)

0.346(0.056)

0.128 (0.016)

0.206(0.034)

RandomForest (396)

0.228 (0.063)

0.336(0.066)

0.193 (0.057)

0.283(0.047)

0.122 (0.020)

0.159(0.019)

RandomForest (257)

0.208 (0.067)

0.336(0.066)

0.195 (0.091)

0.283(0.047)

0.119 (0.091)

0.159(0.019)

RaptorX-Roll (358)

0.146 (0.034)

0.412(0.076)

0.164 (0.028)

0.344(0.058)

0.105 (0.026)

0.271(0.041)

PLCT (332)

0.116 (0.027)

0.329(0.070)

0.095(0.016)

0.275(0.050)

0.073(0.008)

0.173(0.024)

SVM (81)

0.087 (0.03)

0.253(0.061)

0.090 (0.03)

0.243(0.051)

0.069 (0.020)

0.162(0.023)

1d-rec. NN (125)

0.075 (0.022)

0.338(0.076)

0.067 (0.017)

0.290(0.055)

0.046 (0.009)

0.183(0.031)

  1. Accuracies for the top L/10, L/5 and L medium and long range contact predictions for 13 hard targets. L is the length of the protein. Estimates for standard error are provided in parenthesises.