Skip to contents

TWAS Knockoff main function

Source: R/twasknockoff.R
TwasKnockoff.Rd

TWAS Knockoff main function

Usage

TwasKnockoff(
  snpidx,
  ye,
  Xe,
  summarystat,
  Xp,
  removemethod = "lasso",
  simu = FALSE,
  reduced = TRUE,
  lambda_r = 0.1,
  correlation = "improved",
  nrep = 10,
  ts = "lasso",
  appr = "sdp",
  yep_true = NULL,
  gene_fdr = "genesnp"
)

Arguments

snpidx

A list of numeric vectors, storing the indices of cis-variants for each candidate gene in the risk region.

ye

A list of numeric vectors, storing the gene expression levels for each gene in the eQTL study.

Xe

A list of matrices, storing the cis-genotype matrices for each gene in the eQTL study.

summarystat

Summary statistics for cis-SNPs in the risk region.

Xp

The cis-genotype matrix from GWAS data or reference panel for the risk region.

removemethod

If 'lasso', remove the significant variants (eQTLs) detected in the gene expression prediction model.

simu

If TRUE, print the correlation between predicted gene expression levels and true gene expression levels given as input via yep_true.

reduced

If TRUE, only consider the cis-SNPs within each gene; if FALSE, consider all cis-SNPs in the risk region.

lambda_r

Parameter for the regularization of correlation matrix.

correlation

If 'improved', apply the improved correlation matrix estimation via bootstrap sample.

nrep

Number of bootstrap samples (including the original copy).

ts

Test statistic for knockoffs, selected from 'marginal','susie','lasso','lasso.approx.lambda','squared.zscore'.

appr

Approximation method for knockoffs.

yep_true

A list of true gene expression levels for GWAS study, only required when simu = TRUE.

gene_fdr

If 'gene': only use genes for FDR control; If 'genesnp': use all genetic elements (including genes and cis-SNPs) for FDR control.

Value

A list of three elements, the first element contains test statistics for all genetic elements, the second element contains q-values and knockoff statistics similar to GhostKnockoff, the third element contains SNPs removed from the knockoff procedure (i.e., significant SNPs in gene expression prediction models)