IFish: predicting the pathogenicity of human nonsynonymous variants using gene-specific/family-specific attributes and classifiers.

ABSTRACT: Accurate prediction of the pathogenicity of genomic variants, especially nonsynonymous single nucleotide variants (nsSNVs), is essential in biomedical research and clinical genetics. Most current prediction methods build a generic classifier for all genes. However, different genes and gene families have different features. We investigated whether gene-specific and family-specific customized classifiers could improve prediction accuracy. Customized gene-specific and family-specific attributes were selected with AIC, BIC, and LASSO, and Support Vector Machine classifiers were generated for 254 genes and 152 gene families, covering a total of 5,985 genes. Our results showed that the customized attributes reflected key features of the genes and gene families, and the customized classifiers ach

SUBMITTER: Wang M 

PROVIDER: S-EPMC4985647 | biostudies-literature | 2016 Aug

REPOSITORIES: biostudies-literature