PFStatistics

PFStatistics is a personalized variant testing and selection framework that addresses how a variant impact on a phenotype varies continuously along some heterogeneity features, such as genetic ancestry.

Installation

To install from GitHub:

# If devtools is not installed, first install it
install.packages("devtools")

# Then install PFStatistics from GitHub
devtools::install_github("juliefrwang/PFStatistics")

Usage

Loading the Package

After installation, load the package with:

library(PFStatistics)

Main Functions Overview

get_importance_matrices

This main function performs the core analysis by calculating the importance matrices for selected genetic variants.

Usage:

results <- get_importance_matrices(
  genetic_variants = genetic_variants,
  genetic_variants_knockoff = genetic_variants_knockoff,
  additional_covariates = pcs,
  Z = eur, 
  y = y,
  n_folds = 10,  
  FDR_rate = 0.1  
)

Arguments:

• genetic_variants: Matrix of original genetic variants (SNPs), with dimensions n x p where n represents the number of samples (individuals), and p represents the number of SNPs.
• genetic_variants_knockoff: Matrix of knockoff genetic variants, structured identically to genetic_variants with dimensions n x p, where each column is a knockoff version of the corresponding SNP in genetic_variants.
• additional_covariates: Matrix of additional covariates, with dimensions n x c, where c represents the number of covariates, it can be NULL.
• Z: Heterogeneity variable (e.g., estimated genetic ancestry or PCs), dimensions n x m, where m is the number of heterogeneity variables
• y: Outcome variable, a vector of length n, representing the response variable (e.g., disease status or continuous phenotype) for each individual.
• n_folds: Number of folds for cross-validation.
• FDR_rate: Target false discovery rate for feature selection.

The function returns a list containing:

• coefs: Extracted model coefficients.
• scaled_selection_matrix: A matrix indicating scaled selection of SNPs.
• selection_matrix: A binary matrix indicating SNP selection.
• W_statistic_matrix: W-statistic matrix for SNPs.

generate_knockoff_data

This function generates knockoff variables for the given genetic variant matrix. Knockoff variables are used in the Lasso model to control false discovery rates.

Usage:

knockoff_matrix <- generate_knockoff_data(genetic_variants_matrix)

Arguments:

• genetic_variants_matrix: Matrix containing SNP data where rows represent individuals (samples) ane columns represent genetic variants.

Example

Below is an example of using PFStatistics to analyze SNP data:

# Load SNP
snp_filepath <- "snp_data.csv"
snp_data <- read.csv(snp_filepath)

# Extract genetic_variants matrix and generate knockoff data
pcs <- snp_data[, c("PC1", "PC2", "PC3", "PC4")]
eur <- snp_data$EUR
y <- snp_data$AD
genetic_variants <- snp_data[, grepl("chr", colnames(snp_data))]  # Extract columns with "chr" in their names
genetic_variants_knockoff <- generate_knockoff_data(genetic_variants)

# Perform importance calculation
results <- get_importance_matrices(
  genetic_variants = genetic_variants,
  genetic_variants_knockoff = genetic_variants_knockoff,
  additional_covariates = pcs,
  Z = eur, # or other variables like 'pcs'
  y = y,
  n_folds = 10,
  FDR_rate = 0.1
)

Contact