A unified approach to genotype imputation and haplotype-phase inference for large data sets of trios and unrelated individuals

Abstract

We present methods for imputing data for ungenotyped markers and for inferring haplotype phase in large data sets of unrelated individuals and parent-offspring trios. Our methods make use of known haplotype phase when it is available, and our methods are computationally efficient so that the full information in large reference panels with thousands of individuals is utilized. We demonstrate that substantial gains in imputation accuracy accrue with increasingly large reference panel sizes, particularly when imputing low-frequency variants, and that unphased reference panels can provide highly accurate genotype imputation. We place our methodology in a unified framework that enables the simultaneous use of unphased and phased data from trios and unrelated individuals in a single analysis. For unrelated individuals, our imputation methods produce well-calibrated posterior genotype probabilities and highly accurate allele-frequency estimates. For trios, our haplotype-inference method is four orders of magnitude faster than the gold-standard PHASE program and has excellent accuracy. Our methods enable genotype imputation to be performed with unphased trio or unrelated reference panels, thus accounting for haplotype-phase uncertainty in the reference panel. We present a useful measure of imputation accuracy, allelic R(2), and show that this measure can be estimated accurately from posterior genotype probabilities. Our methods are implemented in version 3.0 of the BEAGLE software package.

Figure 1

Calibration of Posterior Genotype Probabilities Genotypes for chromosome 1 markers on the Illumina 550K chip, but not the Affymetrix 500K chip, were imputed with a phased reference panel of 60 individuals (HapMap CEU panel) in a sample of 1088 individuals genotyped on the Affymetrix 500K chip. Imputed genotypes are divided into bins according to their posterior genotype probability. The proportion of imputed genotypes that are consistent with the Illumina genotype are given for each bin. The line is the set of points with equal posterior genotype probability and accuracy rate.

Figure 2

Imputation Accuracy and Reference Panel Sample Size Genotypes for markers on the Illumina 550K chip, but not the Affymetrix 500K chip, were imputed in a sample of 188 individuals with five different reference panels: 60 phased individuals and 100, 300, 600, and 1200 unphased individuals. For each reference panel, the proportion of imputed markers whose allelic R² (see Material and Methods) exceeds each threshold is given. The allelic R² for each imputed marker is calculated with the assumption that the Illumina genotypes are the true genotypes.

Figure 3

Median Allelic R² and Minor-Allele Frequency Genotypes for markers on the Illumina 550K chip, but not the Affymetrix 500K chip, were imputed with two different reference panels in a sample of 1088 individuals genotyped on the Affymetrix 500K chip. For each minor-allele frequency, x = 0.01, 0.02, …, 0.5, the median allelic R² for imputed markers with minor-allele frequency between x − 0.01 and x + 0.01 is plotted.

Figure 4

Allelic Test p values for SNPs Associated with Disease in the WTCCC Study Allelic test p values were computed from data for approximately 2000 cases and approximately 2500 controls genotyped with the Affymetrix 500K chip. Two reference panels were used for imputing data: 300 unphased individuals genotyped on both the Affymetric 500K and Illumina 550K chips and 60 phased individuals from the HapMap CEU panel. For each marker of interest, p values were calculated for the original genotype data and for the imputed data obtained from each reference panel. The imputed data for each marker of interest was obtained after masking the genotype data for the marker in the sample. The allelic test was a two-sample t test of the estimated allele dosage in each individual. Left panel: p values for 15 markers (outside the MHC) that have minor-allele frequency >0.10 in controls, that show the strongest association (p < 5 × 10⁻⁷) on an allelic or genotypic test in the WTCCC study, and that have evidence of association in replication studies. Right panel: p values for nine markers with minor-allele frequency between 0.06 and 0.10 in controls that were reported to show moderate or strong association (p < 10⁻⁵) on an allelic or genotypic test in the WTCCC study. One marker (rs6679677) that is associated with two diseases (rheumatoid arthritis and type 1 diabetes) is repeated.

Figure 5

Building the BEAGLE HMM (A) Building level l+1 from level l. The first step is merging. In this example, states S₁, S₂, and S₄ are merged, and states S₃ and S₅ are merged. After merging, haplotype clusters are split on the basis of the allele at marker l+1. All haplotypes in states S₇ and S₁₀ have allele 1 at this marker, whereas all haplotypes in states S₈, S₉, and S₁₁ have allele 2. (B) Transition probabilities between the states at the two levels. All transitions with nonzero probabilities are shown. Transitions with the same probability have the same pattern on the arrow shaft.