Is the gene-environment interaction paradigm relevant to genome-wide studies? The case of education and body mass index.

This figure consists of three conceptual line graphs illustrating different gene-environment interaction models: Diathesis-Stress, Differential Susceptibility, and Social Push. Each plot maps Body Mass Index (y-axis) against Genotype (x-axis), comparing a "Healthy environment" (dashed line) and an "Unhealthy environment" (solid line). The models show varying relationships, ranging from a genotype effect present only in unhealthy environments (Diathesis-Stress) to opposing effects based on environment (Differential Susceptibility) and a genotype effect present only in healthy environments (Social Push).

This figure consists of a QQ plot (left) and a Manhattan plot (right) showing GWGEI-GxE estimates for body mass index. The QQ plot compares observed versus expected -log10 values, with data points closely following the diagonal line. The Manhattan plot displays -log10 observed values across genomic coordinates (chromosomes 1-22 and X), with a horizontal threshold line at approximately 2.

This figure consists of six scatter plots with linear regression lines showing the relationship between six different genetic variants (rsIDs) and BMI, stratified by education level ("No college" vs. "College"). The x-axis represents the genotype (0, 1, 2) and the y-axis represents BMI. Each plot includes the Minor Allele Frequency (MAF) and statistical values for genetic effect (PV: G), gene-environment interaction (GxE), and $r^2$.

This figure consists of four scatter plots comparing genome-wide association parameter estimates and p-values between college graduates and noncollege graduates (top row, "True") and two simulated environments (bottom row, "Simulated"). The left panels plot estimates for both groups, dividing the data into eight numbered quadrants via dashed lines, while the right panels plot $-\log_{10}(p)$ values. In all plots, the majority of data points cluster near the origin or along the axes, with a single highlighted black point visible in the "True" data plots.

This figure consists of two scatter plots comparing p-values from GWGEI models. The left panel is a comparison plot showing $-\log_{10}(p)$ values for "Original" models on the x-axis versus "No Pop Strat" models on the y-axis, displaying a dense cluster of points with some dispersion. The right panel is a Quantile-Quantile (Q-Q) plot comparing $-\log_{10}(\text{Observed})$ p-values against $-\log_{10}(\text{Expected})$ p-values, where the data points closely follow the diagonal identity line before deviating upward at higher values.