Table 1 Summary of the features of the predictors used in this study
From: Performance of Web tools for predicting changes in protein stability caused by mutations
Name of the predictor | Web server address | Input data | Type of approach | Dataset used for testing | References |
|---|---|---|---|---|---|
PopMuSiC | http://www.dezyme.com/en/Software (accessible to registered users only; free of charge for academic users) | Protein structure | Linear combination of statistical potentials whose coefficients depend on the solvent accessibility of the mutated residue, correcting the bias toward destabilizing mutations imposing physical symmetries under inverse mutations | Direct dataset: S2648 dataset, made of 2648 different point-mutations (602 stabilizing and 2046 destabilizing) across 131 globular proteins with experimentally determined structures and impact on protein stability, extracted from ProTherm. Inverse dataset: constructed from all inverse mutations belonging to the direct dataset by modelling each inverse mutant with Modeller: 2648 mutations on 2648 different 3D structures | [12] |
INPS-3D | Protein sequence (INPS) or structure (INPS3D) | Support vector regression trained on descriptors encoding mutation type (in particular, substitution score, Hydrophobicity score, mutability index of native residue, molecular weights of native and mutant residues) and evolutionary information (INPS). Addition of structural features such as relative solvent accessibility of native residue and local energy difference calculated by a contact potential (INPS3D) | (i) S2648 dataset; (ii) a subset of S2648 used as blind test set comprising 351 variations in 60 proteins; (iii) a dataset of 42 variations within the DNA-binding domain of the tumor suppressor protein P53 | [18] | |
DUET | Protein structure | Meta-predictor: consensus prediction of two complementary approaches developed by the same research group (mCSM and SDM) and obtained by combining the results using SVM | S2648 dataset. In particular: the training set comprises 2297 mutations randomly selected from the S2648 dataset; the test set is composed of the other 351 non-redundant mutations | [17] | |
DynaMut | Protein structure | Meta-predictor: consensus among different predictors (Bio3D, ENCoM and DUET). The first two predictors are based on normal mode analysis of the conformational variability, the last one is a consensus method based on two approaches calculating statistical potentials | Same as DUET | [20] | |
MAESTROweb | Protein structure | Predictor based on a multi-agent machine learning system estimation, which provides also high-throughput scanning for multi-point mutations and a specific mode for the prediction of stabilizing disulfide bonds | 7 different datasets (4 for single point mutations—including S2648 dataset—for a total of 6688 mutations; 1 for multipoint mutations for a total of 479 mutations; 2 for mutations involving disulfide bonds, for a total of 90 disulfide bonds), extracted by ProTherm and by the datasets used to develop PoPMuSic, IMutant2.0 and AutoMute (see ref. 6 for the description of these tools) | [19] |