Skip to main content

Table 4 Performance of Prethermut and other computational methods on the S-dataset

From: Predicting changes in protein thermostability brought about by single- or multi-site mutations

Method r Q2 (%) na
CC/PBSA 0.56 78.6 478
EGAD 0.59 71.0 1065
FoldX 0.5 69.5 1200
Hunter 0.45 69.4 1594
I-Mutant2.0 0.54 77.5 933
Rosetta 0.26 73.4 1913
Combining method 0.64 80.8 407
Prethermut (RF)b 0.72 78.6 2156
Prethermut (SVM)c 0.70 83.2 2156
  1. See Methods for definitions of overall accuracy (Q2) and Pearson correlation coefficient (r). The prediction results of CC/PBSA, EGAD, FoldX, Hunter, I-Mutant 2.0, Rosetta, and Combining method were obtained from Potapov et al. [2]. an is the number of mutant proteins for which the method correctly predicted the change in thermostability. bThe number of trees in the Random forests (RF) method is 10000. The results were obtained by a 10-fold cross validation on the S-dataset. cThe parameters for the support vector machine (SVM) method are gamma (g) = 2, cost (c) = 4, and the weight for the positive samples (w) = 5. The results were obtained by a 10-fold cross validation on the S-dataset.