Bayesian group Lasso for nonparametric varying-coefficient models with application to functional genome-wide association studies

Jiahan Li, Zhong Wang, Runze Li, Rongling WuThe Annals of Applied Statistics 2015 · DOI · arXiv

Introduction

Genome-wide association studies (GWAS) have revolutionized our understanding of the genetic basis of complex traits. However, most GWAS treat phenotypic traits as static snapshots, ignoring the fact that biological processes are inherently dynamic. Traits such as Body Mass Index (BMI), blood pressure, and hormone levels fluctuate over time, influenced by developmental stages and environmental factors.

To capture these temporal dynamics, researchers have proposed "functional GWAS" (fGWAS). This paper introduces a high-dimensional varying-coefficient framework that treats genetic effects not as fixed constants, but as continuous functions of time. By integrating a Bayesian group lasso approach, the authors provide a robust method for identifying significant single-nucleotide polymorphisms (SNPs) and estimating their time-varying influences on longitudinal traits.

The Statistical Challenge of Dynamic Traits

Analyzing longitudinal genetic data presents three significant statistical hurdles:

  1. High Dimensionality ($p \gg n$): GWAS typically involve hundreds of thousands of SNPs (predictors) but only a few thousand participants, making standard regression models mathematically intractable.
  2. Functional Nature of Effects: The influence of a specific gene may wax and wane over a person's lifespan. A static model might average these effects, potentially missing a gene that is highly active at age 20 but dormant at age 60.
  3. Irregular Sampling: Real-world longitudinal data is rarely uniform. Subjects in studies like the Framingham Heart Study are often measured at irregular intervals, resulting in sparse and unbalanced data points.

Methodology: The fGWAS Framework

The authors address these challenges through a high-dimensional varying-coefficient model. The core of their toolkit includes:

Varying-Coefficient Modeling

Instead of a scalar effect size, the model represents the genetic effect of a SNP as a function $\beta(t)$. To make these functions computationally manageable, the authors approximate them using Legendre polynomials. This transformation converts the functional estimation problem into a group-based coefficient estimation problem.

Bayesian Group Lasso

The primary innovation lies in the application of the Bayesian group lasso. Because each SNP is represented by a set (or "group") of polynomial coefficients, the model must decide whether to retain or discard the entire group. The group lasso encourages "group sparsity," effectively performing variable selection by shrinking the coefficients of non-significant SNPs to zero.

Unlike frequentist lasso methods, the Bayesian approach utilizes Markov Chain Monte Carlo (MCMC) algorithms (specifically Gibbs sampling). This allows the researchers to estimate not only the point values of genetic effects but also their credible intervals, providing a clear measure of statistical uncertainty. Furthermore, the model incorporates an AR(1) covariance structure to account for the inherent correlation between repeated measurements of the same individual.

Empirical Evaluation and Real-World Application

Simulation Studies

The researchers first validated the model using extensive simulations involving 3,000 SNPs and 600–800 individuals. The Bayesian group lasso demonstrated superior performance compared to univariate fGWAS, functional principal component analysis (PCA), and simple slope models. It achieved a higher success rate in identifying true causal SNPs while maintaining a near-zero false-positive rate, even in the presence of linkage disequilibrium (where genetic variants are inherited together).

The Framingham Heart Study (FHS)

To demonstrate real-world utility, the authors analyzed BMI data from the Framingham Heart Study. The dataset included 865 participants with repeated measurements taken between the ages of 29 and 61, screened against approximately 33,000 SNPs.

The fGWAS model identified 24 significant SNPs associated with age-specific changes in BMI. Crucially, several of these SNPs are located near well-documented candidate genes for obesity and metabolic health, including FTO, PPARG, and CDKAL1. This biological validation suggests that the model is successfully capturing meaningful genetic signals that govern the trajectory of human growth and aging.

Insights and Future Directions

The fGWAS framework offers a powerful lens for developmental genetics. By treating the genetic architecture as a "movie" rather than a "still," it reveals how certain variants influence the rate of change in a trait. A key advantage of the Bayesian framework is its ability to perform "automatic tuning"; the parameters controlling the sparsity of the model are estimated directly from the data, removing the need for computationally expensive cross-validation.

However, the model's reliance on MCMC makes it computationally intensive. While it handles thousands of SNPs effectively, scaling to the millions of SNPs found in modern whole-genome arrays would likely require a preliminary screening step or further algorithmic optimization. Future work may also expand the model to incorporate gene-gene and gene-environment interactions, further refining our understanding of the dynamic interplay between nature and nurture.

Summary updated

Abstract

Although genome-wide association studies (GWAS) have proven powerful for comprehending the genetic architecture of complex traits, they are challenged by a high dimension of single-nucleotide polymorphisms (SNPs) as predictors, the presence of complex environmental factors, and longitudinal or functional natures of many complex traits or diseases. To address these challenges, we propose a high-dimensional varying-coefficient model for incorporating functional aspects of phenotypic traits into GWAS to formulate a so-called functional GWAS or <i>f</i>GWAS. Bayesian group lasso and the associated MCMC algorithms are developed to identify significant SNPs and estimate how they affect longitudinal traits through time-varying genetic actions. The model is generalized to analyze the genetic control of complex traits using subject-specific sparse longitudinal data. The statistical properties of the new model are investigated through simulation studies. We use the new model to analyze a real GWAS data set from the Framingham Heart Study, leading to the identification of several significant SNPs associated with age-specific changes of body mass index. The <i>f</i>GWAS model, equipped with Bayesian group lassso, will provide a useful tool for genetic and developmental analysis of complex traits or diseases.
BibTeX
@article{li2015bayesian,
  author = {Jiahan Li and Zhong Wang and Runze Li and Rongling Wu},
  title = {{Bayesian group Lasso for nonparametric varying-coefficient models with application to functional genome-wide association studies}},
  journal = {The Annals of Applied Statistics},
  year = {2015},
  doi = {10.1214/15-aoas808},
}
Figure