Optimizing phenotype scale improves genetic analyses in large-scale biobanks
Large-scale biobanks have enabled increasingly complicated genetic analyses across thousands of phenotypes. However, studies rarely consider the appropriateness of the phenotype scale, which can drastically affect inference on genetic architecture. Here we introduce SIQReg, which learns a data-driven phenotype scale by minimizing heterogeneity across phenotype quantiles. Across 25 UK Biobank complex traits, SIQReg rejects the default scale for 24. Generally, SIQReg scales lie between default and logarithmic, indicating that traits are neither purely additive nor purely multiplicative. SIQReg improves both nonadditive and additive genetic analyses. It eliminates most nonadditive signals, including 97% of variance quantitative trait loci, indicating that they may be scale-induced, while preserving biologically plausible nonadditive signals. It simultaneously improves power for additive signals, increasing identification of genome-wide association loci and transcriptome-wide association genes by 11% and 13%, respectively, improving polygenic prediction accuracy by 10%, and identifying 50% more high-risk individuals. These gains are replicated across ancestry groups. Our results establish SIQReg as a principled phenotype scale transformation approach that improves genetic analyses of complex traits.
Authors
- Manuela Costantino (ORCID: https://orcid.org/0000-0003-1886-3173)
- Zhenhong Huang
- Andy Dahl
Institutions
- University of Chicago (US)
Publication Details
- Journal
- Nature Genetics
- Published
- 2026-11-19
- DOI
- https://doi.org/10.1038/s41588-026-02826-1
- Citations
- 3
- Primary Topic
- Genetic Associations and Epidemiology
- Type
- article
- Field-Weighted Citation Impact
- 5.13
Funders
- University of Pennsylvania
- Georgia Clinical and Translational Science Alliance
- National Institutes of Health
- Directorate for Biological Sciences
- National Institute of General Medical Sciences
- Common Fund