Data release for "Markedly divergent performance of variant annotation methods for gene-level association testing"

<p dir="ltr">This is the data release for "Markedly divergent performance of variant annotation methods for gene-level association testing" (<a href="doi.org/10.1186/s12864-026-13379-2" target="_blank" rel="noreferrer">Irudayanathan and Aguirre, et. al., 2026</a>). This release consists of four files in <code>.parquet</code> format, which we provide to the research community so as to enable (1) reproducibility of our work and (2) further analysis of these data.</p><p dir="ltr">One file (<code>.variants.parquet</code>) contains information on variants and their annotations, scores, and labels as computed using the machine-learning tools considered in our study.</p><p dir="ltr">The remaining files contain summary statistics from gene association tests, which were computed using variant labels and gene annotations from the variant file.

The columns of these files are largely consistent with the output format of <a href="rgcgithub.github.io/regenie/" target="_blank" rel="noreferrer">REGENIE</a> (cf. <a href="doi.org/10.1038/s41588-021-00870-7" target="_blank" rel="noreferrer">Mbatchou et. al. 2021</a>), which was used to perform all of our genotype-phenotype analysis. One file (<code>.sumstats.parquet</code>) contains results from our primary analysis of data from the UK Biobank study (cf. <a href="doi.org/10.1038/s41586-018-0579-z" target="_blank" rel="noreferrer">Bycroft et. al., 2018</a>).