A-G.C.07: Efficient Population Structure-Adjusted Logistic Regression for Genome-wide Association Interaction Studies via Clustered Covariates
Authors
Institute of Clinical Molecular Biology, Kiel University and University Medical Center Schleswig-Holstein
Lars Wienbrandt
Institute of Clinical Molecular Biology, Kiel University and University Medical Center Schleswig-Holstein
David Ellinghaus
Institute of Clinical Molecular Biology, Kiel University and University Medical Center Schleswig-Holstein
Keywords
Logistic Regression, Covariates, K-Means, Clustering, Epistasis, GWAS, GWAIS, Genome-wide Association Interaction Studies , Bioinformatics
[sponser-meet-now-chat][/sponser-meet-now-chat]
Genome-wide association interaction studies (GWAIS) provide a robust framework for detecting epistasis, yet they present significant computational challenges due to the extensive search space. Logistic regression remains the standard method for testing statistical interactions; however, covariates that account for population structure, typically derived from principal component analysis (PCA), are often excluded to improve computational efficiency. This exclusion may result in inflated test statistics and biased effect estimates.
This work introduces a novel approximation that combines clustered proxy-covariates with contingency tables to substantially reduce the computational complexity of covariate-adjusted logistic regression for epistasis detection. Rather than including per-sample PCA covariates, population structure is summarized using a limited number of proxy-covariates. This approach reduces runtime complexity from O(N I) to O(N + IK), where N denotes the number of samples, I the number of regression iterations, and K the number of proxy-covariates, while maintaining statistical accuracy relative to the full covariate-adjusted model, as measured by mean relative error (MRE).
The proposed method was evaluated on a real-world case-control genome-wide association study (GWAS) dataset with hospitalized COVID-19 patients and healthy controls. The clustered proxy-covariates reduced the MRE by up to 68% compared to logistic regression without covariates, and achieved speedups of up to 92-fold with 5 clusters relative to logistic regression with per-sample PCA covariates. These findings indicate that covariate clustering enables rapid and statistically meaningful GWAIS analyses, thereby making population-structure-adjusted interaction testing feasible at the genome-wide scale.
Co-authors: Lars Wienbrandt, David Ellinghaus
Contact Attendee
Warning: Attempt to read property "user_email" on string in /home/1276969.cloudwaysapps.com/ydbgzhdjeq/public_html/wp-content/plugins/my-conference-now/functions.php on line 2826