Ordered multinomial regression for genetic association analysis of ordinal phenotypes at Biobank scale

Abstract

Logistic regression is the primary analysis tool for binary traits in genome‐wide association studies (GWAS). Multinomial regression extends logistic regression to multiple categories. However, many phenotypes more naturally take ordered, discrete values. Examples include (a) subtypes defined from multiple sources of clinical information and (b) derived phenotypes generated by specific phenotyping algorithms for electronic health records (EHR). GWAS of ordinal traits have been problematic. Dichotomizing can lead to a range of arbitrary cutoff values, generating inconsistent, hard to interpret results. Using multinomial regression ignores trait value hierarchy and potentially loses power. Treating ordinal data as quantitative can lead to misleading inference. To address these issues, we analyze ordinal traits with an ordered, multinomial model. This approach increases power and leads to more interpretable results. We derive efficient algorithms for computing test statistics, making ordinal trait GWAS computationally practical for Biobank scale data. Our method is available as a Julia package OrdinalGWAS.jl. Application to a COPDGene study confirms previously found signals based on binary case–control status, but with more significance. Additionally, we demonstrate the capability of our package to run on UK Biobank data by analyzing hypertension as an ordinal trait.

Genome-wide association study of self-reported walking pace suggests beneficial effects of brisk walking on health and survival

“…The linear scale makes strong assumptions about the distances between the categories of self-reported walking pace. Whilst recently developed ordinal logistic regression methods have been applied to non-imputed data at UK Biobank scale 38 , they are not yet computationally tractable on densely imputed GWAS datasets. Analysing ordered categorical variables on the linear scale proves problematic when interpreting SNP effect sizes, SNP-heritability and causal effect estimates in MR. We converted heritability estimates from the observed scale to the liability scale, which is more interpretable as it models self-reported walking pace as a continuous trait.…”

“…We used a sample of 373,414 unrelated individuals, such that no pair are related to 3rd degree or above, corresponding to a KING kinship coefficient 50 of <0.044. We fitted both linear and ordinal logistic models with covariates for age, sex, genotyping array and 20 principal components using PLINK1.9 51 for the linear model and the Julia package OrdinalGWAS.jl 38 for the ordinal logistic model.…”

Shared genetic influences on resting-state functional networks of the brain

“…Interesting significant gene-wide associations also included FHL5, a gene previously associated with migraine 47,48 , spatial memory 49 , and cerebral blood flow 50 , and SMYD3, which previously came up in GWASs of cognitive ability 51,52 , suicide attempt 53 , and bipolar disorder 54 . However, the genes retrieved by our gene-wide analyses were not all related to traits exclusively relevant to the brain, but also to cardiovascular 55,56 , metabolic 57,58 , and drug response traits 59,60 . BOLD amplitude, being a blood-based measure, may also be susceptible to genetic effects affecting blood-related traits that are not necessarily specific to the brain.…”

A Genome-Wide Association Study of Novel Genetic Variants Associated With Anthropometric Traits in Koreans

“…Furthermore, HGMA1 and C6orf106 have been reported to show a genetic association with body fat ratios in Caucasians ( Rask-Andersen et al, 2019 ), and both GRM4 and CUX2 were associated with the γ-GT-catalyzed reaction in excessive alcohol consumption ( Chen et al, 2020 ). Additionally, the CUX2 gene has been investigated in genome association studies related to serum uric acid levels, coronary artery disease, and hypertension ( Cho et al, 2020 ; German et al, 2020 ; Matsunaga et al, 2020 ). However, to the best of our knowledge, the association with anthropometric trait (especially the WHR) GWAS signals has never been reported previously.…”