Predicting mendelian disease-causing non-synonymous single nucleotide variants in exome sequencing studies

Area of Science:

• Genomics
• Human Genetics
• Computational Biology

Background:

• Exome sequencing is crucial for identifying Mendelian disease-causing variants.
• Minor allele frequency (MAF) filtering and functional prediction methods are standard for variant identification.
• Combining prediction methods may enhance accuracy.

Purpose of the Study:

• To develop a logit model for combining functional prediction methods to calculate the probability of rare variants being pathogenic.
• To assess the predictive power of seven methods, including the proposed logit model, for pathogenic non-synonymous single nucleotide variants (nsSNVs) after MAF filtering.
• To estimate the burden of pathogenic rare nsSNVs in the general population.

Main Methods:

• Development of a logit model to integrate outputs from multiple functional prediction tools (SIFT, PolyPhen2, CONDEL, etc.).
• Evaluation of the predictive performance of individual methods and the combined logit model on known pathogenic nsSNVs.
• Assessment of prediction performance in the context of post-MAF filtering data from exome sequencing.

Main Results:

• The proposed logit model, integrating multiple prediction methods, demonstrated superior performance compared to individual methods in identifying pathogenic nsSNVs.
• The logit model could not differentiate between autosomal dominant and autosomal recessive disease-causing mutations.
• An estimated 5% of rare nsSNVs are pathogenic, with individuals carrying approximately 22 pathogenic alleles.

Conclusions:

• A logit model effectively combines functional prediction tools to improve the accuracy of identifying pathogenic rare variants from exome sequencing data.
• The developed model provides a more reliable estimation of pathogenic variant burden in individuals.
• Further research is needed to distinguish between dominant and recessive pathogenic variants.