Fig. 4.

Comparison of levels of reference bias for different coverage levels. Input data for imputation in which the evaluation individuals were downsampled to coverages ranging between 0.1 and 2.0× was generated (displayed in different colors), and the difference in reference affinity between HQ and resulting imputed genotypes estimated, with negative values indicating increased reference affinity of the imputed genotypes compared to the HQ genotypes. Imputation was performed using configuration 2 (Table 1), with no filter on posterior genotype probability imposed on the imputed genotypes. Results are shown for all markers, aggregated into 50 MAF bins, and averaged over the evaluation individuals.