Accurate prediction of stability changes in protein mutants by combining machine learning with structure based computational mutagenesis

Masso, Majid; Vaisman, Iosif I.

doi:10.1093/bioinformatics/btn353

Cited by 159 publications

(142 citation statements)

References 39 publications

Supporting

Mentioning

141

Contrasting

Order By: Relevance

“…AUTO-MUTE [Masso and Vaisman, 2008] (http://proteins. gmu.edu/automute/AUTO-MUTE.html) uses a four-body, knowledge-based, statistical contact potential.…”

Section: Prediction Methodsmentioning

confidence: 99%

“…Machine-learning methods [Capriotti et al, 2005[Capriotti et al, , 2008Cheng et al, 2006;Dosztanyi et al, 1997Dosztanyi et al, , 2003Shen et al, 2008] are first trained using examples of proteins and their mutants for which the DDGs have been experimentally measured. Recently a combination of these approaches has been developed [Masso and Vaisman, 2008].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Performance of protein stability predictors

2010

View full text Add to dashboard Cite

Stability is a fundamental property affecting function, activity, and regulation of biomolecules. Stability changes are often found for mutated proteins involved in diseases. Stability predictors computationally predict protein-stability changes caused by mutations. We performed a systematic analysis of 11 online stability predictors' performances. These predictors are CUPSAT, Dmutant, FoldX, I-Mutant2.0, two versions of IMutant3.0 (sequence and structure versions), MultiMutate, MUpro, SCide, Scpred, and SRide. As input, 1,784 single mutations found in 80 proteins were used, and these mutations did not include those used for training. The programs' performances were also assessed according to where the mutations were found in the proteins, that is, in secondary structures and on the surface or in the core of a protein, and according to protein structure type. The extents to which the mutations altered the occupied volumes at the residue sites and the charge interactions were also characterized. The predictions of all programs were in line with the experimental data. I-Mutant3.0 (utilizing structural information), Dmutant, and FoldX were the most reliable predictors. The stability-center predictors performed with similar accuracy. However, at best, the predictions were only moderately accurate ($60%) and significantly better tools would be needed for routine analysis of mutation effects. Hum Mutat 31:675-684,

show abstract

“…AUTO-MUTE [Masso and Vaisman, 2008] (http://proteins. gmu.edu/automute/AUTO-MUTE.html) uses a four-body, knowledge-based, statistical contact potential.…”

Section: Prediction Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Performance of protein stability predictors

2010

View full text Add to dashboard Cite

show abstract

“…Stability changes upon single-site variants were estimated using I-Mutant 3.0 (Capriotti et al, 2008), Eris (Yin et al, 2007), and AutoMute (Masso & Vaisman, 2008). Polyphen (Ramensky et al, 2002), SNPs3D (Yue et al, 2006), PMut (Ferrer-Costa et al, 2005) and SNAP (Bromberg & Rost, 2007) were applied to predict potentially deleterious effects of the new variants.…”

Section: In Silico Analysismentioning

confidence: 99%

Identification and In Silico Analysis of Novel von Hippel-Lindau (VHL) Gene Variants from a Large Population

Leonardi

Martella

Tosatto

et al. 2011

Annals of Human Genetics

View full text Add to dashboard Cite

SummaryMutational inactivation of the VHL gene is the cause of von Hippel-Lindau (VHL) disease, an autosomal dominant hereditary cancer syndrome predisposing to haemangioblastomas, pheochromocytomas and clear-cell renal carcinomas. The gene product (pVHL) functions as an adapter in cellular processes including cell growth and apoptosis. VHL mutation analysis was carried out in 426 unrelated subjects with phenotypes ranging from VHL syndrome, to isolated VHL-related tumours that could represent the first manifestation of the disease. A total of 111 individuals were found to carry alterations, with large deletions representing 40% of the variants. Eighteen of the 95 detected variants were novel, seemingly disease-causing mutations; their pathogenic role has been evaluated in silico for effects on protein folding and interactions. Putative regions of interaction between pVHL and proteins involved in common pathways have been identified, assessing possible implications for the presence of mutations in these regions. All new variants predicted to truncate or cause complete pVHL loss of structure were associated with phenotypes consistent with VHL type 1. Seven of the new amino acid substitutions are disease-causing mutations, one is a neutral variant, whereas the results for two remain ambiguous. Our combined molecular and in silico approach for the evaluation of putative disease-causing mutations contributes to the interpretation of the potential pathogenicity of these novel variants.

show abstract

“…These changes are often calculated using molecular dynamics and a force field energy function (Potapov et al, 2009). However, empirical energy functions and other methods such as statistical potentials (Dehouck et al, 2011(Dehouck et al, , 2009) and machine learning algorithms (Yue et al, 2005;Masso and Vaisman, 2008;Chen et al, 2009;Dehouck et al, 2009;Ferrada and Melo, 2009) have also been employed. Structural descriptors such as solvent accessibility and secondary structure propensity as well as residue-residue interaction networks have also been used with limited success (Li et al, 2011;Martin et al, 2011;Wang et al, 2012;Yang et al, 2013).…”

Section: Introductionmentioning

confidence: 99%

Principal Component Analysis of Binding Energies for Single-Point Mutants of hT2R16 Bound to an Agonist Correlate with Experimental Mutant Cell Response

Chen

Willick²,

Ruckel

et al. 2015

Journal of Computational Biology

View full text Add to dashboard Cite

Directed evolution is a technique that enables the identification of mutants of a particular protein that carry a desired property by successive rounds of random mutagenesis, screening, and selection. This technique has many applications, including the development of G proteincoupled receptor-based biosensors and designer drugs for personalized medicine. Although effective, directed evolution is not without challenges and can greatly benefit from the development of computational techniques to predict the functional outcome of single-point amino acid substitutions. In this article, we describe a molecular dynamics-based approach to predict the effects of single amino acid substitutions on agonist binding (salicin) to a human bitter taste receptor (hT2R16). An experimentally determined functional map of single-point amino acid substitutions was used to validate the whole-protein molecular dynamics-based predictive functions. Molecular docking was used to construct a wild-type agonist-receptor complex, providing a starting structure for single-point substitution simulations. The effects of each single amino acid substitution in the functional response of the receptor to its agonist were estimated using three binding energy schemes with increasing inclusion of solvation effects. We show that molecular docking combined with molecular mechanics simulations of single-point mutants of the agonist-receptor complex accurately predicts the functional outcome of single amino acid substitutions in a human bitter taste receptor.

show abstract

Accurate prediction of stability changes in protein mutants by combining machine learning with structure based computational mutagenesis

Abstract: A web server with supporting documentation is available at http://proteins.gmu.edu/automute.

Cited by 159 publications

References 39 publications

Performance of protein stability predictors

Performance of protein stability predictors

Identification and In Silico Analysis of Novel von Hippel-Lindau (VHL) Gene Variants from a Large Population

Principal Component Analysis of Binding Energies for Single-Point Mutants of hT2R16 Bound to an Agonist Correlate with Experimental Mutant Cell Response

Contact Info

Product

Resources

About