Human leucocyte antigen class I and II imputation in a multiracial population

Area of Science:

• Genetics
• Immunology
• Bioinformatics

Background:

• Human leucocyte antigen (HLA) genes are vital for immune responses and autoimmunity.
• Traditional HLA genotyping is costly and labor-intensive, limiting research.
• Existing imputation algorithms often lack accuracy in diverse ancestries.

Purpose of the Study:

• To evaluate the performance of two HLA imputation algorithms, SNP2HLA and HIBAG.
• To assess imputation accuracy in a multiracial population.
• To determine the reliability of HIBAG for diverse genetic studies.

Main Methods:

• Compared SNP2HLA and HIBAG using 1587 women with gold-standard HLA genotyping.
• Utilized 80% training and 20% testing data split for initial comparison.
• Performed five 10-fold cross-validation procedures with ancestry delineation for HIBAG.

Main Results:

• HIBAG demonstrated equal or superior accuracy compared to SNP2HLA.
• Overall HIBAG imputation accuracy reached 89%, with gene-specific accuracies ranging from 83% to 94%.
• Accuracy varied by ancestry group, being highest in African and lowest in Hispanic individuals.

Conclusions:

• HIBAG is a viable tool for HLA imputation in multiracial populations.
• The algorithm provides posterior accuracy estimates, enabling analysis of high-confidence subsets.
• Further refinement may be needed for specific gene/ancestry combinations to enhance imputation accuracy.