Preprint proposes empirical multiple-testing thresholds for HLA association studies using whole-genome sequencing data

A preprint posted to bioRxiv addresses a practical gap in the methodology of HLA association studies: unlike genome-wide association studies of single-nucleotide variants, for which the field has broadly converged on a genome-wide significance threshold of p < 5 × 10⁻⁸, there is no established consensus for controlling the multiple-testing burden when interrogating classical HLA alleles resolved from whole-genome sequencing data.

The nine classical HLA genes (HLA-A, -B, -C, -DRB1, -DQA1, -DQB1, -DPA1, -DPB1, and -DRB3/4/5) are highly polymorphic and in extensive linkage disequilibrium with one another, meaning that the effective number of independent tests differs substantially from a naïve count of alleles. The authors use population-scale sequencing datasets spanning multiple genetic ancestry groups to estimate empirically how many independent tests are actually being performed, and derive ancestry-stratified thresholds accordingly.

The work is directly relevant to researchers conducting HLA-disease association studies — particularly for autoimmune and infectious-disease phenotypes, where HLA effects are often the strongest genetic signals — who need principled methods to balance false-positive control against statistical power. Because this is a preprint and has not yet completed peer review, the specific threshold values and ancestry-group estimates should be treated as preliminary. Readers should consult the final peer-reviewed publication once available.