Variable Prioritization in Nonlinear Black Box Methods: A Genetic Association Case Study

TL;DR

This paper introduces the RATE measure for variable importance in nonlinear models, especially in genetic studies, providing interpretable insights into predictor significance and covarying relationships.

Contribution

The paper presents a novel, interpretable importance measure called RATE for nonlinear models, applicable beyond Bayesian Gaussian processes, to improve variable prioritization in genetic data.

Findings

RATE improves variable prioritization in nonlinear models.

Application of RATE explains the superior predictive accuracy of nonlinear models.

Simulations and real data demonstrate RATE's effectiveness in genetic association studies.

Abstract

The central aim in this paper is to address variable selection questions in nonlinear and nonparametric regression. Motivated by statistical genetics, where nonlinear interactions are of particular interest, we introduce a novel and interpretable way to summarize the relative importance of predictor variables. Methodologically, we develop the "RelATive cEntrality" (RATE) measure to prioritize candidate genetic variants that are not just marginally important, but whose associations also stem from significant covarying relationships with other variants in the data. We illustrate RATE through Bayesian Gaussian process regression, but the methodological innovations apply to other "black box" methods. It is known that nonlinear models often exhibit greater predictive accuracy than linear models, particularly for phenotypes generated by complex genetic architectures. With detailed simulations and two real data association mapping studies, we show that applying RATE enables an explanation for this improved performance.