Adaptive Robust Method for Multi-trait Genome-wide Association Studies

Abstract: Genome-wide association studies (GWAS) over the past decade have identified thousands of genetic variants associated with human traits or diseases. However, the heritability of many traits remains largely unexplained. Traditional single-trait analysis methods are overly conservative, while multi-trait methods improve statistical power by integrating association evidence from multiple traits. GWAS summary statistics are usually publicly available, so methods using only summary statistics have greater potential for use. To address the inconsistent performance, low computational efficiency, and numerical issues when considering a large number of traits in developed multi-trait analysis methods, we propose an adaptive Fisher method for multi-trait analysis of summary statistics (MTAFS), which is computationally efficient and statistically robust.

Research Background: Genome-wide association studies (GWAS) have played an important role in studying the correlation between genetic variants and complex diseases. However, when a genetic variant is associated with multiple traits, single-trait analysis may lead to loss of statistical power. Given this, developing methods capable of jointly analyzing multiple traits has become increasingly urgent.

Research Source: This study was conducted jointly by Qiaolan Deng, Chi Song, and Shili Lin, affiliated with the Division of Biostatistics, College of Public Health, and the Department of Statistics, College of Arts and Sciences, at The Ohio State University. The research findings were published in the European Journal of Human Genetics (2024), Volume 32, pages 681-690.

Research Details: a) The research process details an adaptive Fisher method for multi-trait analysis of summary statistics (MTAFS). This process includes eigendecomposition of Z-scores for decorrelation, calculation of P-values for each individual trait, and use of the Cauchy method to combine joint evidence from multiple trait analyses.

b) Research results show that MTAFS performs robustly in various background settings, controls type I errors, effectively handles a large number of traits, and demonstrates advantages compared to existing multi-trait methods.

c) Conclusions and research significance elaborate on MTAFS’s progress in statistical and computational efficiency, as well as its potential advantages in explaining associations between specific traits and genetic variants.

d) Research highlights focus on the novelty of MTAFS, aiming to address some key issues in multi-trait analysis, such as efficiency, robustness, and the ability to handle a large number of traits.

Importance and Value: MTAFS provides a new tool for analyzing multi-trait correlations in GWAS. Its research value lies not only in scientific insights into the genetic mechanisms of complex traits but also in potential contributions to future personalized medicine and the prediction and treatment of complex diseases. Through application to the UK Biobank’s brain imaging-derived phenotypes (IDPs) dataset, the method has demonstrated its effectiveness in practical research.

---

This paper not only contributes a new data analysis method to the field of genomic research but also confirms the advantages of multi-trait analysis methods in statistical power and computational efficiency in large-scale GWAS studies. As part of an open-source R package, MTAFS provides researchers with an effective tool to easily implement the method and apply it to their own datasets.