Gene-Environment Interaction in the Era of Precision Medicine - Filling the Potholes Rather Than Starting to Build a New Road.

José M Álvarez-Castro

doi:10.3389/fgene.2020.00921

Abstract

Gene–environment interaction is a key part of evolutionary biology, animal, and plant breeding, and a number of health sciences, like epidemiology and precision medicine. However, bottlenecks in models of gene–environment interaction have recently been made manifest, particularly in the field of medicine and, consequently, specific improvements have been explicitly requested—namely, an implementation of gene–environment interaction satisfactorily disentangled from gene–environment correlation. The present paper meets those demands by providing mathematical developments that implement classical models of genetic effects and bring them up to date with the prospects current available data bestow. These developments are shown to overcome the limitations of previous proposals through the analysis of illustrative examples on disease susceptibility, with special attention paid to precision medicine. Indeed, a number of misconceptions about the application of models of genetic/environmental effects to precision medicine are here identified and clarified. The theory here provided is argued to strengthen, in particular, the methodology required for high-precision characterization of strain virulence in the study of the COVID-19 pandemic.

Highlights

Scientific progress is often accompanied with expectations beyond objective appraisal
Interactions are known to encrypt the map where a pursued genetic architecture could be spotted. This is known to occur because interactions of any kind may make lower level effects vanish under a certain genetic/environmental composition of a population or of an experimental sample
The commendable review by Malosetti et al (2013) on models of gene–environment interaction in the context of plant breeding reasonably recommend to adhere to a strategy where effects are inspected sequentially—as they are in expressions (1–3) above, but it oversteps the mark when proposing a conditional sequential procedure, by claiming that “dominance effects should be tested conditioned on the additive effects present in the model.”

Summary

INTRODUCTION

Scientific progress is often accompanied with expectations beyond objective appraisal. Further on, Li et al (2019) provide a probabilistic approach based on a Bayesian framework to hierarchically model gene– environment interaction, leading to a population-dependent index, C, called the genetic coefficient of the disease (at a population)—“a large C indicates large distinguishability of case genomes from control genomes.” They illustrate the performance of the proposed methodology using a built-up example in which the disease susceptibility is by default very low (0.01) and it significantly increases due to either environmental (exposure) or genetic (risk allele) factors or both, to 0.4, 0.5, and 0.9, respectively. This is as expected under the lack of interplay between gene and environment (i.e., no interaction and no correlation). Such an advantage can hereafter be applied to more complex real cases of interest undergoing less intuitive behaviors

DISCUSSION

DATA AVAILABILITY STATEMENT