Modeling coverage gaps in haplotype frequencies via Bayesian inference to improve stem cell donor selection

Area of Science:

• Genetics
• Computational Biology
• Immunogenetics

Background:

• Accurate genotype imputation is crucial for genetic studies and clinical applications, especially in highly polymorphic regions.
• Existing imputation methods struggle with rare and unobserved haplotypes, leading to errors in applications like stem cell donor matching.
• Human Leukocyte Antigen (HLA) gene imputation is critical for unrelated stem cell transplantation, where novel haplotypes pose a significant challenge.

Purpose of the Study:

• To develop and evaluate a novel statistical method for extending haplotype frequency distributions.
• To improve the accuracy of genotype imputation in highly polymorphic regions, specifically for Human Leukocyte Antigen (HLA) genes.
• To address the challenge of unobserved haplotypes in large donor registries for stem cell transplantation.

Main Methods:

• Applied a Bayesian inference methodology to extend haplotype frequency distributions.
• Utilized a model where new haplotypes are generated through the recombination of observed alleles.
• Validated the method using five-locus HLA frequency data from the National Marrow Donor Program registry.

Main Results:

• The proposed Bayesian method significantly improves frequency distribution estimates compared to existing methods.
• Demonstrated substantial enhancement in imputation accuracy for highly polymorphic regions.
• The methodology effectively addresses the issue of unobserved and rare haplotypes.

Conclusions:

• The developed Bayesian inference framework offers a significant advancement for genotype imputation accuracy.
• This approach has direct applications in improving Human Leukocyte Antigen (HLA) matching for stem cell transplantation.
• The statistical inference model can benefit disease association studies and variant imputation involving rare alleles.