Knowledge-based computational mutagenesis for predicting the disease potential of human non-synonymous single...

Area of Science:

• Genetics and Bioinformatics
• Structural Biology
• Computational Biology

Background:

• Genetic variations, including single nucleotide polymorphisms (SNPs), are linked to heritable diseases.
• Non-synonymous SNPs (nsSNPs) alter protein sequences, potentially affecting protein function and causing disease.
• Understanding nsSNP impact requires analyzing structural changes within proteins.

Purpose of the Study:

• To computationally predict whether human nsSNPs are neutral or disease-related.
• To investigate the relationship between protein structural changes and nsSNP pathogenicity.
• To develop a novel classification method for nsSNP impact assessment.

Main Methods:

• Utilized a dataset of 1790 human nsSNPs mapped to 243 protein structures.
• Employed a computational mutagenesis approach using a statistical contact potential to quantify environmental perturbations.
• Generated vector representations for nsSNPs using structural, sequence, and protein features.
• Applied a random forest supervised classification algorithm.

Main Results:

• Achieved 76% cross-validation accuracy in classifying nsSNPs.
• Demonstrated that structural change data are effective predictors of nsSNP pathogenicity.
• The developed classifier performs comparably to methods using larger datasets.
• Novel attributes provide an orthogonal approach to existing techniques.

Conclusions:

• Computational analysis of structural perturbations offers a viable method for predicting nsSNP disease association.
• The developed model provides an accurate and efficient tool for assessing the impact of genetic variations.
• This approach complements existing methods for disease-related SNP identification.