A New Calling Procedure for Illumina Beadar Ray Data

• Activity Number: 613
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Computing
• Author(s): Gengxin Li*+
• Keywords: Dirichlet Process ; Gaussian Mixture Model ; Single Nucleotide Polymorphisms ; Rare Variants ; Genotype

Abstract

Accurate genotype calling for high throughput Illumina data is a key step to explore more genetic information on a large scale variants for genome wide association studies. Many popular calling algorithms use mixture models with fixed components to fast and efficiently infer genotypes, but mixture models with fixed components are mostly restricted to infer genotypes for common SNPs where their minor allele frequencies are large. For rare variants, developing a mixture model with a model selection procedure for explicitly calling genotypes is still an open question. To improve the call accuracy and the quality of genotypes on SNP microarrays, a two-model calling procedure is proposed, named (GMM+DP/R), to accurately infer genotypes for the Illumina chip data, and Dirichlet Process Gaussian Mixture Model with reference SNP selection is developed in this calling procedure to integrate the model selection procedure with the predominance of the population-based strategy and the SNP-based approach. More desirable performance is shown through the comparison result between GMM+DP/R and other powerful calling algorithms.