HIBAG--HLA genotype imputation with attribute bagging.

Area of Science:

• Immunogenetics
• Computational Biology
• Genomic Medicine

Background:

• High-resolution human leukocyte antigen (HLA) genotyping is crucial for understanding diseases and adverse drug reactions linked to the major histocompatibility complex (MHC).
• Whole-genome SNP typing or sequencing for HLA alleles can be prohibitively expensive for large-scale studies.
• Existing HLA imputation methods often have usability limitations or demand extensive training data.

Purpose of the Study:

• To introduce HIBAG (HLA Imputation using attribute BAGging), a novel method for imputing high-resolution HLA types from dense SNP genotypes.
• To provide a cost-effective and accessible alternative to direct HLA typing for large cohorts.
• To develop a user-friendly tool that does not require researchers to possess large training datasets.

Main Methods:

• HIBAG employs an ensemble learning approach, averaging posterior probabilities from multiple classifiers trained on bootstrap samples of SNP and HLA data.
• Performance was evaluated using a training set of 2668 European ancestry subjects and independent validation data from the British 1958 birth cohort (approx. 1000 subjects).
• SNP markers common to various Illumina platforms (1M Duo, OmniQuad, OmniExpress, 660K, 550K) were utilized for imputation.

Main Results:

• HIBAG achieved high prediction accuracies for key HLA loci (HLA-A, B, C, DRB1, DQB1), ranging from 92.2% to 98.1%.
• The method demonstrated competitive performance when compared against established imputation tools like HLA*IMP and BEAGLE.
• The HIBAG R package includes pre-fit classifiers for diverse ancestries (European, Asian, Hispanic, African), enhancing its broad applicability.

Conclusions:

• HIBAG offers a robust and accurate solution for HLA imputation from SNP data, facilitating large-scale genetic association studies.
• The method's accessibility, particularly its independence from user-provided training data, significantly lowers the barrier for HLA imputation.
• HIBAG represents a valuable advancement for genomic research in immunogenetics and personalized medicine.