Evaluation of computational techniques for predicting non-synonymous single nucleotide variants pathogenicity.

Area of Science:

• Genetics and Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Non-synonymous Single Nucleotide Variants (nsSNVs) are a significant factor in human genetic diseases, altering protein function and potentially leading to disease.
• Numerous computational methods exist to predict the functional impact of nsSNVs and classify them as pathogenic or neutral.

Purpose of the Study:

• To systematically assess the performance of eight established computational tools (FATHMM, SIFT, Provean, iFish, Mutation Assessor, PANTHER, SNAP2, and PON-P2) in predicting the pathogenicity of nsSNVs.
• To develop and evaluate a novel meta-classifier combining top-performing individual tools to enhance prediction accuracy.

Main Methods:

• Utilized a curated dataset, VaribenchSelectedPure, comprising 2144 pathogenic and 3777 neutral variants.
• Evaluated eight individual nsSNV prediction tools based on specificity, sensitivity, AUC, and accuracy on both the full dataset and a random sample.
• Developed a meta-classifier (CSTJ48) by integrating the outputs of FATHMM, iFish, and Mutation Assessor.

Main Results:

• Individual tools showed varying performance, with FATHMM exhibiting the highest specificity (83.75%) and sensitivity (94.13%) on the whole dataset.
• The meta-classifier CSTJ48 significantly outperformed all individual tools, achieving 96.33% specificity, 86.07% sensitivity, 91.20% AUC, and 91.89% accuracy.
• FATHMM demonstrated the best performance among the eight individual tools evaluated.

Conclusions:

• The performance of individual nsSNV prediction tools varies, highlighting the need for careful selection and validation.
• The developed meta-classifier CSTJ48 offers a significant improvement in accurately classifying nsSNVs, showing great potential for disease-associated variant identification.
• Computational tools, especially ensemble methods like CSTJ48, are crucial for understanding the genetic basis of diseases caused by nsSNVs.