PLSKO

R package and source codes used in paper, 'PLSKO: a robust knockoff generator to control false discovery rate in omics variable selection'(https://www.biorxiv.org/content/10.1101/2024.08.06.606935v1)

Installation

You can install the development version of the package from GitHub using the following code:

# install.packages("devtools")
devtools::install_github("guannan-yang/PLSKO/PLSKO", quiet = TRUE)

If warnings of installing dependencies appears, please install the dependencies manually by running the following code:

# install.packages(c("knockoff","progress", "parallel", "doParallel", "foreach"))
#
# if (!require("BiocManager", quietly = TRUE))
#     install.packages("BiocManager")
# BiocManager::install("mixOmics")

Vignette

View the vignette

Knockoff Steps and Functions Pipeline

Basic Input

Input: predictor matrix X
Dimensions: n x p
Input: response vector y
Length of n
Input: Target False Discovery Rate (FDR) Level
Threshold: q ∈ (0, 1]
Output: knockoff.result object with Selected variables

Main functions included in this package

				Pipeline Functions Provided^
Main Steps of Knockoff framework	Function	Auxiliary Function	*Output (Class)*	`plsko_filter()`, `plsAKO()`	`ko_filter()`, `AKO_with_KO()`
Step 1: Knockoff Variable Generation	`plsko()`	`r_criterion()` (for `ncomp` estimation)	A `n x p` matrix of knockoff variables	✔️	(Bring your own knockoff vars)
Step 2: Importance Score Calculation (`W`)	(import from pkg `Knockoff`)	-	A vector of `p` or a `n_ko x p` matrix	✔️	✔️
Step 3: Knockoff Filtering and Variable Selection	`KO_with_W()` `AKO_with_W()`	-	A list (class "knockoff.result" or "AKO.result") with components: `statistic`, `selected`, `ako.selected`	✔️	✔️

View the Knockoff Framework Diagram

Additional Notes:

^The workflow includes generating knockoff variables, importance score calculation, and filtering.