Skip to main content

Table 2 Evaluation of protein model quality assessment approaches

From: UQlust: combining profile hashing with linear-time ranking for efficient clustering and analysis of big macromolecular data

Method CASP10 TASSER
PconsD 0.68 / 0.43 4.3 / 0.46
uQlust:1D-CA-CM 0.66 / 0.38 4.2 / 0.46
uQlust:1D-SS-SA 0.67 / 0.40 4.3 / 0.41
ClusCo (10) 0.68 / 0.37 3.2 / 0.49
Pleiades (10) 0.67 / 0.38 3.1 / 0.45
uQlust: Hash (10,60) 0.76 / 0.52 3.5 / 0.44
uQlust: Rpart (10,60) 0.75 / 0.56 3.3 / 0.42
uQlust:Tree 0.71 / 0.46 2.9 / 0.47
  1. Average MaxSub similarity score between top ranking and best models (left), and fraction of good models (right) are reported for both CASP and TASSER targets. The fraction of good models is defined as the fraction of targets with the top ranking model less than 0.2 MaxSub score from the best model for CASP, and less than 2 Ang RMSD for TASSER. Centroids of the 5 largest (out of K = 10) clusters are considered for clustering methods, and F = 60% of data is used for uQlust