HISAT-genotype repeatability and agreement with SNP2HLA.

To assess the repeatability of HISAT-genotype in imputing HLA types, for each HLA type in each of the participants with multiple profiled time points, the difference between HLA type dosage at each time point and the mean dosage was calculated (13). Nine comparisons from four participants gave a difference in dosage of >50%. Of these, eight resulted from a clear outlying time point in each participant. These time points were therefore excluded, and the difference between each time point and the mean was recalculated. For each HLA type, intraclass correlation coefficients (one way, single measurement) were calculated using the irrNA package to assess agreement.

For participants with multiple time points HLA typed by HISAT-genotype, the median HLA type dosage was taken. HLA dosages were then rounded to 50%. For 71 participants both genotyping and RNA-sequencing data were available. To assess the agreement between HLA types imputed by HISAT-genotype and SNP2HLA, weighted Cohen’s kappa was calculated for each HLA type using the irr R package. HLA type dosage was considered to be an ordinal variable (100% >50% >0%). HLA types with a weighted Cohen’s kappa of <0.4, both at 2-digit and at 4-digit resolution, were excluded from association analysis.