MAPPIN: a method for annotating, predicting pathogenicity and mode of inheritance for nonsynonymous variants. Issue 18 (25th August 2017)

• Record Type: Journal Article
• Authors: Gosalia, Nehal; Economides, Aris N.; Dewey, Frederick E.; Balasubramanian, Suganthi
• Is Part Of: Nucleic acids research. Volume 45:Issue 18(2017)
• Page Start: 10393
• Page End: 10402
• Publication Date: 2017-08-25
• Subjects: Nucleic acids -- Periodicals; Molecular biology -- Periodicals; 572.805
• DOI: 10.1093/nar/gkx730
• Languages: English
• ISSNs: 0305-1048

Abstract

Abstract: Nonsynonymous single nucleotide variants (nsSNVs) constitute about 50% of known disease-causing mutations and understanding their functional impact is an area of active research. Existing algorithms predict pathogenicity of nsSNVs; however, they are unable to differentiate heterozygous, dominant disease-causing variants from heterozygous carrier variants that lead to disease only in the homozygous state. Here, we present MAPPIN (M ethod for A nnotating, P redicting P athogenicity, and mode of I nheritance for N onsynonymous variants), a prediction method which utilizes a random forest algorithm to distinguish between nsSNVs with dominant, recessive, and benign effects. We apply MAPPIN to a set of Mendelian disease-causing mutations and accurately predict pathogenicity for all mutations. Furthermore, MAPPIN predicts mode of inheritance correctly for 70.3% of nsSNVs. MAPPIN also correctly predicts pathogenicity for 87.3% of mutations from the Deciphering Developmental Disorders Study with a 78.5% accuracy for mode of inheritance. When tested on a larger collection of mutations from the Human Gene Mutation Database, MAPPIN is able to significantly discriminate between mutations in known dominant and recessive genes. Finally, we demonstrate that MAPPIN outperforms CADD and Eigen in predicting disease inheritance modes for all validation datasets. To our knowledge, MAPPIN is the first nsSNV pathogenicity prediction algorithm that provides mode of inheritance predictions, adding another layer of information for variant prioritization.