Efficient phasing and imputation of low-coverage sequencing data using large reference panels

Abstract

Low-coverage whole genome sequencing followed by imputation has been proposed as a cost-effective genotyping approach for disease and population genetics studies. However, its competitiveness against SNP arrays is undermined as current imputation methods are computationally expensive and unable to leverage large reference panels. Here, we describe a method, GLIMPSE, for phasing and imputation of low-coverage sequencing datasets from modern reference panels. We demonstrate its remarkable performance across different coverages and human populations. It achieves imputation of a full genome for less than $1, outperforming existing methods by orders of magnitude, with an increased accuracy of more than 20% at rare variants. We also show that 1x coverage enables effective association studies and is better suited than dense SNP arrays to access the impact of rare variations. Overall, this study demonstrates the promising potential of low-coverage imputation and suggests a paradigm shift in the design of future genomic studies.

Characterization of a haplotype- reference panel for genotyping by low-pass sequencing in Swiss Large White pigs

“…We considered the 70 key ancestor animals as a reference to call genotypes from low-pass sequencing data (1.11-fold) of genetically similar pigs. In agreement with previous studies in human and cattle populations, the genotyping accuracy from the low-pass sequencing data was very high [6,14,56]. Moreover, the low-pass sequencing-derived genomic relationship coe cients were highly correlated with those obtained using microarray genotyping.…”

“…To compile the reference haplotypes, we retained 22,618,811 biallelic autosomal SNP that were polymorphic (minor allele count ≥ 1) among the 70 key ancestor pigs. Following the approach proposed by Rubinacci et al [6], we used the mpileup and call commands of BCFtools [67] to calculate genotype likelihoods at the 22,618,811 polymorphic sites in the 175 low-pass sequenced and reference-aligned samples. Subsequently, we applied the phasing and imputation algorithm implemented in GLIMPSE_phase [6] to re ne the BCFtools-derived genotype calls using the previously established haplotype reference panel.…”

“…This method is particularly useful in species for which dense microarray-derived genotypes are not available. Recent investigations [6,13,14] suggest that a sequencing coverage less than 1-fold is su cient to accurately infer genotypes at known loci -provided an informative haplotype reference panel is available.…”

“…The availability of sequence variant genotypes from key ancestor animals enables imputing sequencelevel genotypes for animals that had been genotyped at lower density [6][7][8]. In livestock populations that are routinely genotyped using 60K genotyping arrays, sequence variant genotypes are typically imputed using stepwise imputation [9].…”

Characterization of a haplotype-reference panel for genotyping by low-pass sequencing in Swiss Large White pigs

“…The availability of sequence variant genotypes from key ancestor animals enables imputing sequence-level genotypes for animals that had been genotyped at lower density [6–8]. In livestock populations that are routinely genotyped using 60K genotyping arrays, sequence variant genotypes are typically imputed using stepwise imputation [9].…”