Example 1: Using SEAGLE with .txt Input Files

In SEAGLE: Scalable Exact Algorithm for Large-Scale Set-Based Gene-Environment Interaction Tests

This tutorial demonstrates how to use the SEAGLE package when the user inputs ${\bf y}$, ${\bf X}$, ${\bf E}$, and ${\bf G}$ from .txt files. We'll begin by loading the SEAGLE package.

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(SEAGLE)

If you have your own files ready to read in for ${\bf y}$, ${\bf X}$, ${\bf E}$, and ${\bf G}$, you can read them into R using the read.csv() command.

As an example, we've included y.txt, X.txt, E.txt, and G.txt files in the extdata folder of this package. The following code loads those files into R so we can use them in this tutorial.

y_loc <- system.file("extdata", "y.txt", package = "SEAGLE")
y <- as.numeric(unlist(read.csv(y_loc)))

X_loc <- system.file("extdata", "X.txt", package = "SEAGLE")
X <- as.matrix(read.csv(X_loc))

E_loc <- system.file("extdata", "E.txt", package = "SEAGLE")
E <- as.numeric(unlist(read.csv(E_loc)))

G_loc <- system.file("extdata", "G.txt", package = "SEAGLE")
G <- as.matrix(read.csv(G_loc))

Now we can input ${\bf y}$, ${\bf X}$, ${\bf E}$, and ${\bf G}$ into the prep.SEAGLE function. The intercept = 1 parameter indicates that the first column of ${\bf X}$ is the all ones vector for the intercept.

This preparation procedure formats the input data for the SEAGLE function by checking the dimensions of the input data. It also pre-computes a QR decomposition for $\widetilde{\bf X} = \begin{pmatrix} {\bf 1}{n} & {\bf X} & {\bf E} \end{pmatrix}$, where ${\bf 1}{n}$ denotes the all ones vector of length $n$.

objSEAGLE <- prep.SEAGLE(y=as.matrix(y), X=X, intercept=1, E=E, G=G)

Finally, we'll input the prepared data into the SEAGLE function to compute the score-like test statistic $T$ and its corresponding p-value. The init.tau and init.sigma parameters are the initial values for $\tau$ and $\sigma$ employed in the REML EM algorithm.

res <- SEAGLE(objSEAGLE, init.tau=0.5, init.sigma=0.5)
res$T
res$pv

The score-like test statistic $T$ for the G$\times$E effect and its corresponding p-value can be found in res$T and res$pv, respectively.

SEAGLE documentation built on Nov. 6, 2021, 1:06 a.m.