Reference choice affects positional enrichment of AlphaGenome variant rankings at APOL1-MYH9

Abstract Objective. We examined how the choice of measured chromatin-accessibility reference set affects interpretation of AlphaGenome variant rankings at the APOL1-MYH9 locus. This exploratory evaluation arose from a regulatory-variant prioritisation project and included a beta-globin comparison locus. Results. Among 1,930 common alternate alleles at 1,921 positions, the top 200 variants ranked by the absolute podocyte-labelled differential accessibility score were enriched in whole-kidney peaks (odds ratio 5.98, 95% CI 4.18–8.52) and in a podocyte-inclusive kidney union (8.76, 6.30–12.20). The podocyte estimate exceeded the strongest non-renal estimate numerically under both references, 5.98 against 5.46 for lung and 8.76 against 5.61, with greater separation under the union. We detected no change in the podocyte-minus-comparator enrichment contrast for any of six comparators after partitioning that union, under a conditional peak bootstrap. At the comparison locus the corresponding estimates were 1.44 (0.16–6.56) and 7.54 (3.66–15.31). These observations illustrate reference-dependent interpretation of positional enrichment. They do not establish tissue-specific variant-effect accuracy, equivalent performance between loci, or generalisation beyond training-overlapping data.