Ananda S Datta1, Shili Lin2, Swati Biswas3. 1. Department of Mathematical Sciences, University of Texas at Dallas, Richardson, Texas, USA. 2. Department of Statistics, The Ohio State University, Columbus, Ohio, USA. 3. Department of Mathematical Sciences, University of Texas at Dallas, Richardson, Texas, USA, swati.biswas@utdallas.edu.
Abstract
BACKGROUND: The variants identified in genome-wide association studies account for only a small fraction of disease heritability. A key to this "missing heritability" is believed to be rare variants. Specifically, we focus on rare haplotype variant (rHTV). The existing methods for detecting rHTV are mostly population-based, and as such, are susceptible to population stratification and admixture, leading to an inflated false-positive rate. Family-based methods are more robust in this respect. METHODS: We propose a method for detecting rHTVs associated with quantitative traits called family-based quantitative Bayesian LASSO (famQBL). FamQBL can analyze any type of pedigree and is based on a mixed model framework. We regularize the haplotype effects using Bayesian LASSO and estimate the posterior distributions using Markov chain Monte Carlo methods. RESULTS: We conduct simulation studies, including analyses of Genetic Analysis Workshop 18 simulated data, to study the properties of famQBL and compare with a standard family-based haplotype association test implemented in FBAT (family-based association test) software. We find famQBL to be more powerful than FBAT with well-controlled false-positive rates. We also apply famQBL to the Framingham Heart Study data and detect an rHTV associated with diastolic blood pressure. CONCLUSION: FamQBL can help uncover rHTVs associated with quantitative traits.
BACKGROUND: The variants identified in genome-wide association studies account for only a small fraction of disease heritability. A key to this "missing heritability" is believed to be rare variants. Specifically, we focus on rare haplotype variant (rHTV). The existing methods for detecting rHTV are mostly population-based, and as such, are susceptible to population stratification and admixture, leading to an inflated false-positive rate. Family-based methods are more robust in this respect. METHODS: We propose a method for detecting rHTVs associated with quantitative traits called family-based quantitative Bayesian LASSO (famQBL). FamQBL can analyze any type of pedigree and is based on a mixed model framework. We regularize the haplotype effects using Bayesian LASSO and estimate the posterior distributions using Markov chain Monte Carlo methods. RESULTS: We conduct simulation studies, including analyses of Genetic Analysis Workshop 18 simulated data, to study the properties of famQBL and compare with a standard family-based haplotype association test implemented in FBAT (family-based association test) software. We find famQBL to be more powerful than FBAT with well-controlled false-positive rates. We also apply famQBL to the Framingham Heart Study data and detect an rHTV associated with diastolic blood pressure. CONCLUSION: FamQBL can help uncover rHTVs associated with quantitative traits.
Authors: Daniel J Schaid; Charles M Rowland; David E Tines; Robert M Jacobson; Gregory A Poland Journal: Am J Hum Genet Date: 2001-12-27 Impact factor: 11.025
Authors: Steve Horvath; Xin Xu; Stephen L Lake; Edwin K Silverman; Scott T Weiss; Nan M Laird Journal: Genet Epidemiol Date: 2004-01 Impact factor: 2.135