A comparison of genetic imputation methods using Long Life Family Study genotypes and sequence data with the 1000 Genome reference panel. (2nd February 2020)

• Record Type: Journal Article
• Authors: Kraja, Aldi T.; Daw, E. Warwick; Lenzini, Petra; Wang, Lihua; Lin, Shiow J.; Williams, Christine A.; Wells, Alan B.; Lunetta, Kathryn L.; Murabito, Joanne M.; Sebastiani, Paola; Tosto, Giuseppe; Barral, Sandra; Minster, Ryan L.; Yashin, Anatoly; Perls, Thomas; Province, Michael A.
• Is Part Of: International journal of bioinformatics research and applications. Volume 16:Number 1(2020)
• Page Start: 59
• Page End: 84
• Publication Date: 2020-02-02
• Subjects: genetic imputation -- 1000 Genomes reference -- sequence reference -- MACH software -- MINIMACH software -- SHAPEIT2 software -- IMPUTE2 software -- FCGENE software -- LLFS -- long life family study; Bioinformatics -- Periodicals; 570.285
• Languages: English
• ISSNs: 1744-5485

Abstract

This study compares methods of imputing genetic markers, given a typed GWAS scaffold from the Long Life Family Study (LLFS) and latest reference panel of 1000-Genomes. We examined two programs for pre-phasing haplotypes MACH/SHAPEIT2 and MINIMAC/IMPUTE2 for imputation. SHAPEIT2 is advantageous for haplotype pre-phasing. MINIMAC and IMPUTE2 produced similar imputation quality. We used a 4MB region on chromosome 2 of LLFS and in the Supplement, we compared methods using chromosome 19 data from the Genetic Analysis Workshop-19. IMPUTE2 had the advantage of using two references 1000G and a sequence for a subset of subjects. SHAPEIT2 and IMPUTE2 were used to finalise the full LLFS autosome imputation. In LLFS, 44% of ~80M autosomal imputed variants showed good imputation quality (info ≥ 0.30). Low imputation quality was associated with a predominantly low allele frequency in 1000-Genomes. New emerging large-scale sequences and enhanced imputation methodologies will further improve imputation quality.