GxE

The goal of GxE is to enable easy estimation of gene-environment (GxE) interaction from individual-level polygenic scores (GRS) and outcome phenotypes alone.

Installation

You can install the latest version of GxE from GitHub with:

install.packages( "devtools" )
library( devtools )
install_github( "zkutalik/GRSxE_software", subdir = "Rcode" )

Example

The main function is estimate_gxe, which requires the outcome phenotype of interest y (corrected for any relevant covariates) and the polygenic score, grs. This returns a list containing a separate list for the real data (real_data) and the simulated GRS (fake_grs). Each of these lists contain a vector, coefficients, with the average estimates obtained for each of the 4 parameters, namely alpha1 (linear genetic effect), alpha2 (quadratic genetic effect), beta (environmental effect), and gamma (GxE interaction effect), as well as vectors for the corresponding standard errors (se) and p-values (p) and a matrix with the individual estimates of each bootstrap permutation (individual_estimates).

If simulate_phenotype was used, the returned list fake_phenotype contains similar information for the simulated phenotype, without alpha2 as quadratic genetic effects are accounted for in the phenotype simulation. Additionally, this list contains the skewness and kurtosis values which most closely match the data, the root mean square difference between the real and fake phenotypes (rms_diff), the main linear genetic effect (alp), and the estimates of the genetic effects from linear to seventh power (fY_coefficients).

In addition, the t_real_fgrs contains the t-statistic for the difference between the results from the real and fake GRS.

library(GxE)

gxe  =  estimate_gxe( y, grs )
print( gxe )

In addition, the ukb_estimate_gxe is a helper function to allow easy analysis of UK Biobank data (requires access to a local copy of individual-level UK Biobank genetic and phenotypic data). It allows covariates to be specified and corrected for before estimating GxE (age, age^2, sex, and the top 10 genetic PCs are used as covariates by default).

gxe  =  ukb_gxe_interaction( phenotype_name = '21001-0.0',
                             ukb_filename   = 'uk_biobank/pheno/ukb21067.csv',
                             bgens_path     = 'uk_biobank/imp',
                             snps           = snps,
                             sqc_filename   = 'uk_biobank/geno/ukb_sqc_v2.txt',
                             fam_filename   = 'uk_biobank/plink/ukb1638_cal_chr1_v2_s488366.fam' )
print( gxe )

get_betas_from_neale is another helper function included to automatically extract and prune SNPs and their effect sizes from files as those provided by the Neale Lab.

snps = get_betas_from_neale( neale_filename = '21001_irnt.gwas.imputed_v3.both_sexes.tsv.gz',
                             variants_filename = 'variants.tsv.gz' )