Back to catalogue >>

Reference: Li B., Krishnan V.G., Mort M.E., Xin F., Kamati K.K, Cooper D.N., Mooney S.D., Radivojac P. Automated inference of molecular mechanisms of disease from amino acid substitutions. Bioinformatics (2009) 25 (21) 2744-2750.
Hosted: (http://mutpred.mutdb.org/)

Summary:
MutPred uses a random forest algorithm based on the probabilities of gain or loss of properties relating to many features of protein structure and function.

Methodology:
There are 3 main categories of classification attributes used in the random forest:

1. Protein structure and dynamics:
• Secondary structure
• Solvent accessibility
• Stability
• Intrinsic disorder
• B-factor
• Transmembrane helix
• Coiled-coil structure

2. Predicted functional properties:
• DNA-binding residues
• Catalytic residues
• Calmodulin-binding targets
• Phosphorylation sites
• Methylation sites
• Glycosylation sites
• Ubiquitination sites

3. Amino acid sequence and evolutionary information:
• Score based on SIFT scores of evolutionary conservation

The loss and gain of structural and functional properties are modelled via posterior probabilities. Through assessment of these probabilities, MutPred can predict the molecular cause of disease-associated substitution. The training data comes from the human gene mutation database (HGMD) and neutral polymorphisms from Swiss-Prot.

Input:
The wild-type protein sequence in FASTA format must be pasted and the substitution sites identified. A link to the results will be emailed once available.

Output:
The probability of the mutation being deleterious is reported. The user can apply their own thresholds for significance but P<0.05 is considered appropriate. Any molecular mechanisms that are likely to be disrupted due to the mutation are reported, with corresponding P value.