EVALUATING THE ACCURACY OF GENOTYPE IMPUTATION IN THE MAJOR HISTOCOMPATIBILITY COMPLEX REGION IN SELECTED AFRICAN POPULATIONS

Abstract

Genome wide association studies (GWAS) typically use genotyping arrays to genotype large sets of individuals and thus determine which SNPs are significantly overrepresented in the cases compared to the controls and in this way determine association with disease. Genotyping arrays are cheaper than sequencing but only measure a portion of selected SNPs across the genome. To increase the number of SNPs available, one can use a reference panel of whole genome sequence data from related populations to impute SNPs from those on the array. Some regions in the human genome such as the Major Histocompatibility Complex (MHC), also referred to as the Human Leukocyte Antigen (HLA) region, are highly variable and thus difficult to impute. The HLA region in humans plays important roles in autoimmune and infectious diseases, adaptive and innate immune responses, and adverse responses to organ transplantation. In view of this, it is important to evaluate the accuracy of HLA imputation especially in African populations as they have high diversity, and this has not been extensively studied. The aim of this study was to therefore evaluate the accuracy of HLA imputation in selected African populations. The study sets were selected from the Gambian individuals within the Gambian Genome Variation Project (GGVP). The Illumina Omni 2.5 array and H3Africa array data were inferred from the GGVP datasets using matching markers. The reference datasets were chosen from the 1000 Genomes population (1kg-All), the African sub-population within the 1000 Genomes Project (1kg-Afr), the Gambian sub-population within the 1000 genome project (1kg-Gwd) and the H3Africa reference panel. HLA-A, HLA-B and HLA-C alleles were imputed using HIBAG, SNP2HLA while HLA SNPs were imputed using Minimac4, IMPUTE5 and SNP2HLA imputation tools. The assessment metrics for HLA imputation were concordance rate and squared Pearson correlation coefficient. The most preferable software was HIBAG for HLA alleles imputation and IMPUTE5 vi for HLA SNPs imputation. The 1kg-All reference panel was the best performing reference panel for HLA alleles imputation implying that the reference panel sample size influences HLA alleles imputation. For HLA SNPs imputation, the 1kg-Gwd reference outperformed the other reference panels depicting that population specificity is key when imputing HLA SNPs. The H3Africa array and Illumina Omni 2.5 array performance was comparable for both the HLA alleles and HLA SNPs imputation showing that genotyping arrays have less influence on HLA imputation in African populations. SNPs with low minor allele frequencies (MAF) were imputed less accurately suggesting the need to build new algorithms and larger population specific reference panels with an aim of improving imputation of HLA SNPs with low MAF.