Table 3 Summary of the performance of PDNET-Contact and PDNET-Distance on the 9 CASP13 free-modeling domains (for which the native structures are publicly available) and on the 131 CAMEO hard-targets.

From: A fully open-source framework for deep learning protein real-valued distances

Dataset Methods \(P_{L}\) \(P_{NC}\)
CASP13 FM PDNET-Contact 38.8 38.5
PDNET-Distance 32.3 30.3
Top CASP13 group 45.0 45.7
CAMEO-HARD PDNET-Contact 48.3 45.1
PDNET-Distance 46.7 43.7
trRosetta 48.0
  1. Performance of CASP13’s top group and the trRosetta method on the CAMEO-HARD dataset are provided as a reference. Results of the trRosetta method are copied from the Yang et al.18 since the predicted contacts are not publicly available.