Logistic Bayesian Lasso for Detecting Rare Haplotypes Using Family Trios

• Activity Number: 34
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Meng Wang*+ and Shili Lin
• Companies: The Ohio State University and The Ohio State University
• Keywords: rare variants ; Regularization ; retrospective likelihood ; haplotype

Abstract

Evidences are accumulating that rare variants can explain part of the "missing heritability" in complex traits. Advances in next-generation sequencing techniques have facilitated the identification of rare single nucleotide variants (SNVs), and tatistical methods for detecting rare SNVs from sequencing data have been developed. However, with next-gen sequencing , there still exist technical issues with rare SNVs calling. On the other hand, rare haplotypes can be formed by common SNVs, available from Genome-Wide Association Studies (GWAS). Analysis methods for rare haplotypes remain less explored, especially for family data. Here we propose a retrospective likelihood Bayesian LASSO model for family trios (famLBL). We have found from simulations that for homogeneous population data, famLBL has comparable power with LBL (Biswas and Lin, Biometrics, 2011; DOI: 10.1111/j.1541-0420.2011.01680.x) but is robust to population substructure. Furthermore, famLBL is more powerful for detecting rare variants than the hbat version of FBAT (http://www.biostat.harvard.edu/~fbat/fbat.htm). We will also discuss results for comparing famLBL with other popular rare variants detection methods.