Skip to main content
. 2021 Feb 24;12:1264. doi: 10.1038/s41467-021-21541-5

Fig. 2. Imputation accuracy comparison.

Fig. 2

Accuracies were measured based on the matches in the P-group level. a Prediction accuracy in imputing HLA of the HapMap CEU (N = 88) using the Type 1 Diabetes Genetic Consortium (T1DGC) data (N = 5225) as a reference. b Prediction accuracy in imputing the 1000 Genomes European (EUR) populations (N = 404) using the T1DGC data (N = 5225) as a reference. c Prediction accuracy in cross-validation using the split data from the T1DGC panel, where we used a subset as our target sample (N = 4225) and the rest as a reference (N = 1000). For HIBAG-prefit in (b) and (c), we used the European prefit model. The error bars represent SD.