Prethermut (Predicting changes in protein thermostability brought about by single- or multi-site mutations)
One important task of protein design is the ability to predict protein thermostability changes upon single or multiple site mutations, which could be reflected by free energy changes (¦¤¦¤G).
We propose a method, Prethermut, to identify the effect of single- or multi-site mutations on protein thermostability based on machine learning methods. The input vector of Prethermut considers structural changes and empirical potential energy differences upon protein mutation. Using a 10-fold cross validation test on the M-dataset consisting of 3366 mutants from ProTherm, the classification accuracy of random forests and the regression accuracy of random forest regression were slightly better than support vector machines and support vector regression, whereas the overall accuracy of classification and the Pearson correlation coefficient of regression were 79.2% and 0.72, respectively. Moreover, Prethermut has a propensity for higher performance on mutants with multi-site mutations than those with single-site mutations.
Predicting changes in protein thermostability brought about by single- or multi-site mutations. BMC Bioinformatics. 2010 Jul 2;11(1):370. [PubMed] [PDF]
Jian Tian, Ningfeng Wu*, Xiaoyu Chu and Yunliu Fan
