Fast and accurate imputation of summary statistics enhances evidence of functional enrichment

• DOI: 10.1093/bioinformatics/btu416
• Publication Date: 2014-07-03T00:27:22Z
• Authors: Bogdan Pasaniuc; Noah Zaitlen; Huwenbo Shi; Gaurav Bhatia; Alexander Gusev; Joseph Pickrell; Joel Hirschhorn; David P. Strachan; Nick Patterson; Alkes L. Price
• Keywords: Imputation (statistics); Genome-wide Association Study; Linkage Disequilibrium; 1000 Genomes Project; Statistical power; Genetic Association; Summary statistics

Abstract Motivation: Imputation using external reference panels (e.g. 1000 Genomes) is a widely used approach for increasing power in genome-wide association studies and meta-analysis. Existing hidden Markov models (HMM)-based imputation approaches require individual-level genotypes. Here, we develop new method Gaussian from summary statistics, type of data that becoming available. Results: In simulations Genomes (1000G) data, this recovers 84% (54%) the effective sample size common (>5%) low-frequency (1–5%) variants [increasing to 87% (60%) when linkage disequilibrium information available target samples] versus gold standard 89% (67%) HMM-based imputation, which cannot be applied statistics. Our accounts limited panel, crucial step eliminate false-positive associations, it computationally very fast. As an empirical demonstration, apply our seven case–control phenotypes Wellcome Trust Case Control Consortium (WTCCC) study height British 1958 birth cohort (1958BC). statistics 95% (105%) (as quantified by ratio χ2 statistics) compared with genotypes at 227 (176) published single nucleotide polymorphisms (SNPs) WTCCC (1958BC height) data. addition, publicly large meta-analyses four lipid traits, release imputed 1000G SNPs, could not have been obtained previously methods, demonstrate their accuracy masking subsets We show increases magnitude statistical evidence enrichment genic non-genic loci these as analysis without imputation. Thus, will valuable tool future functional analyses. Availability implementation: Publicly software package http://bogdan.bioinformatics.ucla.edu/software/ . Contact: [e-mail] or [e-mail] Supplementary information: materials are Bioinformatics online.