Pitfalls in performing genome-wide association studies on ratio traits

Abstract

Genome-wide association studies (GWAS) are often performed on ratios composed of a numerator trait divided by a denominator trait. Examples include body mass index (BMI) and the waist-to-hip ratio, among many others. Explicitly or implicitly, the goal of forming the ratio is typically to adjust the numerator for the denominator. While forming ratios may be clinically expedient, there are several important issues with performing GWAS on ratios. Forming a ratio does not “adjust” for the denominator in the sense of holding it constant, and it is unclear whether associations with ratios are attributable to the numerator, the denominator, or both. Here we demonstrate that associations arising in ratio GWAS can be entirely denominator-driven, implying that at least some associations uncovered by ratio GWAS may be due solely to a putative adjustment variable. In a survey of 10 exemplar ratios, we find that the ratio model disagrees with the adjusted model (performing GWAS on the numerator while conditioning on the denominator) at around 1/3 of loci. Using BMI as an example, we show that variants detected by only the ratio model are more strongly associated with the denominator (height), while variants detected by only the adjusted model are more strongly associated with the numerator (weight). Although the adjusted model provides effect sizes with a clearer interpretation, it is susceptible to collider bias. We propose and validate a simple method of correcting for the genetic collider bias via leave-one-chromosome-out polygenic scoring.

The genetic architecture and evolutionary consequences of the human pelvic form

“…We included height as a covariate to directly adjust for differences in body size between individuals and focus on skeletal proportions instead of overall length. We also adjusted for body size differences in two other ways: dividing each phenotype by height to generate a skeletal proportion, and including a leave one- chromosome-out polygenic risk score (PRS) for height as a covariate in the GWAS ( 43 ). GWAS effect sizes using either height as a covariate or height combined with the one-chromosome-out PRS as a covariate were highly correlated (Pearson correlation = 0.99) ( Methods : Adjusting for height correlation in GWAS by adding height as covariate , Fig.…”

“…A major issue in carrying out GWAS for phenotypes such as ours where we would like to control for height is the potential for confounding due to the adjustment. McCaw et al highlight the pitfalls in GWAS of ratio traits and describe ways to reduce this confounding ( 43 ). Following their pipeline, we carried out GWAS, adjusting not only for height but also for leave-one-chromosome-out (LOCO) polygenic scores (PGS) of height.…”

Machine learning across multiple imaging and biomarker modalities in the UK Biobank improves genetic discovery for liver fat accumulation

“…Therefore, we performed principal component analysis (PCA) on the metabolites dataset and used all principal components (PCs; same number as the number of metabolites) as inputs to our ML model (described below). We also included age, sex, and body mass index (BMI) terms - weight (in kilograms), inverse of height (in meters), and inverse of height squared - due to the potential for biased effect estimates from the inclusion of ratio covariates 29 , as inputs to our model.…”

Challenging the current hypothesis that thrombosis is responsible for the post-COVID-19 condition

“…The lower levels of CRP relative to fibrinogen and D-dimer may relate to other processes, such as less detection by individuals of symptoms of infection, and thus, a lower relative CRP as a denominator in their 2 measures could relate to greater or lesser detection or recognition of symptoms. Problems of using ratio traits have recently been discussed elsewhere [ 9 ].…”