Back

Multi-ancestral GWAS with the VA Million Veteran Program enables functional interpretation of rheumatoid arthritis alleles

Sakaue, S.; Yang, D.; Zhang, H.; Posner, D.; Rodriguez, Z.; Love, Z.; Cui, J.; Budu-Aggrey, A.; Ho, Y.-L.; Costa, L.; Monach, P.; Huang, S.; Ishigaki, K.; Melley, C.; Tanukonda, V.; Sangar, R.; Maripuri, M.; Sweet, S. M.; Panickan, V.; McDermott, G.; Hanberg, J. S.; Riley, T.; Laufer, V.; Okada, Y.; Scott, I.; Bridges, S. L.; Baker, J.; VA Million Veteran Program, ; Wilson, P. W.; Gaziano, J. M.; Hong, C.; Verma, A.; Cho, K.; Huffman, J. E.; Cai, T.; Raychaudhuri, S.; Liao, K. P.

2026-04-23 genetic and genomic medicine
10.64898/2026.04.22.26351423 medRxiv
Show abstract

Introductory ParagraphRheumatoid arthritis (RA) is a heritable and common autoimmune condition. To date, most genetic associations were derived from individuals with either European or East Asian ancestries. Here, we applied a multimodal automated phenotyping strategy to define RA and performed a genome-wide association study (GWAS) of RA in the Million Veteran Program (MVP), including underrepresented African American (AFR) and Admixed American (AMR) populations. Meta-analyses with previous RA cohorts identified 152 autosomal genome-wide significant loci, of which 31 were novel. Inclusion of multi-ancestry data dramatically improved fine-mapping resolution. Functional characterization of these loci using single-cell transcriptomic and chromatin data suggested new RA genes such as CHD7 and CD247. We identified underappreciated functional roles of fine-grained immune cell states other than T cells, such as B cell and myeloid cell states. We observed that multi-ancestry polygenic risk scores using our data demonstrated better predictive ability, especially for AFR and AMR populations.

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.