Revisit Population-based and Family-based Genotype Imputation

Abstract

Genome-Wide Association (GWA) with population-based imputation (PBI) has been successful in identifying common variants associated with complex diseases; however, much heritability remains to be explained and low frequency variants (LFV) may contribute. To identify LFV, a study of unrelated individuals may no longer be as efficient as a family study, where rare population variants can be frequent in families. Family-based imputation (FBI) provides an opportunity to evaluate LFV. To compare the performance of PBI and FBI, we conducted extensive simulations, generating genotypes using SeqSIMLA from various reference panels for families. We masked genotype information for variants unavailable in Framingham 550 K GWA genotype data in less informative subjects selected by GIGI-Pick. We implemented IMPUTE2 with duoHMM in SHAPEIT (Impute2_duoHMM) for PBI, MERLIN and GIGI for FBI and PedBLIMP for a hybrid approach. In general, FBI in both MERLIN and GIGI outperformed other approaches with imputation accuracy greater than 0.99 for the squared correlation and imputation quality scores (IQS) especially for LFV, although imputation accuracy from MERLIN depends on pedigree splitting for larger families. PBI performed worst with the exception of good imputation accuracy for common variants when a closely ancestry matched reference is used. In summary, linkage disequilibrium (LD) information from large available genotype resources provides good imputation for common variants with well-selected reference panels without requiring densely sequenced data in family members, while imputation of LFV with FBI benefits more from information on inheritance patterns within families yielding better imputation.

Figure 1

Flow chart of the study design.

Figure 2

Comparison of imputation accuracy with squared correlation and imputation quality score among various imputation strategies under the scenario that all informative people have dense genotypes for family-based imputation and all individuals have sparse genotype as backbone for population-based imputation, using genotypes simulated from haplotype pool 1 (UK10K + 1000 G sample). The left panel, (a,b), of the figure shows the imputation accuracy for the scenario with 1000 families of size 12; the right panel, (c,d), of the figure shows the imputation accuracy for the scenario with 300 families of size 40. For family size of 40, two different pedigree-splitting methods (Merlin_mb18 and Merlin_mb20) are considered based on maxbits of 18 or 20 for the imputation using Merlin. The x-axis in the figures is on a logarithmic scale.

Figure 3

Comparison of imputation accuracy with squared correlation, (a) and imputation quality score, (b) among various imputation strategies under the scenario that all informative people have dense genotypes for family-based imputation and all individuals have sparse genotype as the backbone for population-based imputation, using genotypes simulated from haplotype pool 2 (all UK10K sample). The x-axis in the figures is on a logarithmic scale.

Figure 4

Comparison of imputation accuracy with squared correlation, (a) and imputation quality score, (b) among various population-based imputation strategies under the scenario that all individuals have sparse genotypes as the backbone for population-based imputation, using genotypes simulated from haplotype pool 3 (1/3 of UK10K sample). The x-axis in the figures is on a logarithmic scale.