Table 2 MAE and the precision of the contact prediction method (PDNET-Contact), the distance prediction method (PDNET-Distance), and binned distance prediction method (PDNET-Binned), on the 150 proteins in the test set.

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

Methods \(MAE_{8}\) \(P_{L}\) \(P_{NC}\)
PDNET-Contact 69.5 61.1
PDNET-Distance 4.1 67.5 59.1
PDNET-Binned 3.8 67.8 60.5
DeepCov\(^{{\text {a}}}\) 55.1
DEEPCON 63.4 55.8
  1. All MAEs are in Å, and the \(P_{L}\) and \(P_{NC}\) for the three PDNET methods are shown in percentage. DeepCov and DEEPCON methods’ performances, which are trained on the same dataset, are also reported.
  2. \(^{{\text {a}}}\)As reported in the DeepCov16 paper.