Related Experiment Videos
A novel association test for multiple secondary phenotypes from a case-control GWAS
1Department of Biostatistics and Center for Statistical Genetics, University of Michigan, Ann Arbor, Michigan, United States of America.
Genetic Epidemiology
|April 11, 2017
Summary
Genome-wide association studies (GWASs) often reuse case-control data for secondary traits. A new method, proportional odds model adjusted for propensity score (POM-PS), provides valid genetic association signals, unlike existing approaches.
Area of Science:
- Genetics
- Biostatistics
- Epidemiology
Background:
- Genome-wide association studies (GWASs) commonly use case-control designs to identify SNP associations with complex diseases.
- Existing GWAS datasets collect extensive data on risk factors and traits beyond the primary disease.
- There is a growing trend to re-analyze these datasets for genetic associations with secondary phenotypes.
Purpose of the Study:
- To address the challenge of analyzing secondary phenotypes in case-control GWAS data, where phenotypes may be correlated.
- To evaluate the validity of existing multivariate methods and propose a novel approach that accounts for non-random sampling.
- To ensure accurate Type I error control when secondary traits are associated with both genotypes and disease status.
Main Methods:
- Proposed the proportional odds model adjusted for propensity score (POM-PS) method.
- POM-PS utilizes a proportional odds logistic regression of genotypes on secondary phenotypes.
- Adjusts for the estimated conditional probability of disease status to account for non-random sampling.
Main Results:
- Standard multivariate methods and ad hoc adjustments can lead to substantial Type I error inflation when analyzing secondary traits.
- The effectiveness of using disease status as a covariate is highly dependent on unknown disease mechanisms and causal structures.
- Extensive simulations demonstrated the validity and advantages of POM-PS over existing methods.
- Application to adiposity traits in a type 2 diabetes (T2D) case-control sample (METSIM study) yielded valid association signals exclusively with POM-PS.
Conclusions:
- Existing methods for analyzing secondary phenotypes in case-control GWAS data are prone to inflated Type I errors.
- The proposed POM-PS method offers a statistically sound approach for valid genetic association inference in such scenarios.
- POM-PS successfully identified valid genetic associations for adiposity traits in the METSIM study, highlighting its practical utility.