HLA
Imputation of FinnGen HLA alleles is done for each data freeze by the Finnish Red Cross Blood Service. Analysis team runs association of imputed classical HLA alleles with FinnGen endpoints for every data freeze.
Method
HLA imputation in FinnGen is implemented using R library HIBAG with imputation models trained on Finnish individuals. The Finnish reference data have been genotyped at clinical grade accuracy at 4-digit (i.e. amino acid) level resolution for the HLA genes HLA-A, B, C, DPB1, DQA1, DQB1, DRB1, DRB3, DRB4 and DRB5. HLA imputation of FinnGen data is based on a set of SNPs directly genotyped on the FinnGen array, however, this set is extracted from SNP-imputed data so that approximately the same set of SNPs should be present in all samples regardless of their original genotyping platform.
Output
In addition to allelic non-synonymous variants, HLA-DRB3, DRB4, and DRB5 harbor variants where one or both copies of the gene are not expressed or present. These are denoted as 'ng' while in other cases the allele name is shown as usual.
Furthermore, the algorithm may not always be able to impute an HLA allele to the 4-digit resolution. In that case, the result is shown at 2-digit resolution, e.g. A:01.
The output of HIBAG imputation includes posterior probabilities (PPs) over the reference alleles for each target locus. The imputation result is transformed into a VCF file format, where REF codes for absence of an allele while ALT codes for the presence of an allele. The PP information is incorporated into the VCF GP field. For example, 0|0:0.97,0.03,0 means that both copies of an allele are imputed to be absent at 0.97 PP while the PP for one allele being present is only 0.03 and both alleles being present is 0.
The most recent HLA imputation can be found by looking at the index file in /finngen/shared. We expect to have them incorporated into the core GWAS for DF9.
You can read more details of the HLA imputation process from recently published paper by Jarmo Ritari et al. :
There is also a GitHub repository for the imputation: https://github.com/FRCBS/HLA-imputation
and for the PheWas https://github.com/FRCBS/HLA_PheWAS
International haplotype frequencies can be found from: http://www.allelefrequencies.net/
Analysis team runs association of imputed classical HLA alleles with FinnGen endpoints for every data freeze. Summary stats you can find for example for release 9 from here: /finngen/library-green/finngen_R9/finngen_R9_analysis_data/hla
Last updated