Introduction: Most genome-wide association studies (GWAS) of polygenic traits are based on the assumption that allelic effects are additive. Hundreds of loci associated with type 2 diabetes (T2D) have been identified using this approach, but a significant proportion of heritability remains unexplained. Many genetic variants behave in a recessive fashion, and rare variants of this type can be missed by additive analyses.
Methods: We used a recessive model to conduct a GWAS of T2D in unrelated individuals of European ancestry in the UK Biobank and then meta-analyzed the results with publicly available data from five European-ancestry cohorts known collectively as 70KforT2D. We used approximate conditional analysis to identify distinct secondary signals within genome-wide significant loci.
Results: In total, there were 30,661 cases and 267,948 controls. We identified 54 associated loci and 15 other conditionally independent signals. Six associations were distinct from the set of previously reported signals, including results from larger additive analyses, and three were over a million base pairs away from the closest known locus. Five of the novel variants - located near the genes PELO, ADAMTS6, JAZF1, IL7, and FCHSD2 - were rare, with allele frequencies under 5%, and had large effect sizes, with odds ratios for homozygous carriers ranging from 2.7 to 7.9.
Conclusions: Our results demonstrate that recessive analysis identifies loci missed by additive GWAS, confirming the value of recessive analyses. Some of the variants we detected have large effect sizes and could be useful for patient risk stratification. Collectively they add to our knowledge of the genetic architecture of T2D.
M.J. O’Connor: None. P. Cortes-Sanchez: None. S. Bonàs-Guarch: None. J.B. Cole: None. D. Torrents: None. A. Leong: None. J.C. Florez: Advisory Panel; Self; Doris Duke Charitable Foundation. Other Relationship; Self; Novo Nordisk Inc., Park Street School. J.M. Mercader: None.