Skip to contents

Run multivariate feature-based GSEA using limma and Hallmark gene sets

compute_factor_gsea.Rd

This function fits a multivariate linear model for each gene using all features in features_df as continuous covariates. For each feature, it extracts the moderated t-statistics and p-values, ranks genes, and performs GSEA (fgsea) using the Hallmark gene sets from MSigDB. Optional dotplots for the top enriched pathways can be saved as PDFs.

Usage

compute_factor_gsea(
  RNA.tpm,
  features_df,
  plot_dot = TRUE,
  top_n = 10,
  file_name = NULL,
  width = 8,
  height = 10
)

Arguments

RNA.tpm

A numeric matrix or data frame of gene expression values (genes in rows, samples in columns).

features_df

A data frame of continuous features (samples in rows, features in columns) to be modeled as covariates. Rows must be the same samples, in the same order, as the columns of RNA.tpm (checked by name; an error is raised otherwise).

plot_dot

Logical; if TRUE, generates and saves dotplots of top enriched Hallmark pathways for each feature. Default is TRUE.

top_n

Integer; number of top pathways to display in the dotplot. Default is 10.

file_name

Character; optional suffix for saved PDF files. Default is NULL.

width

Numeric; width of the PDF plot in inches. Default is 8.

height

Numeric; height of the PDF plot in inches. Default is 10.

Value

A list containing:

DE_results

A named list of topTable results for each feature, including logFC, moderated t-statistics, p-values, and adjusted p-values.

GSEA_results

A named list of fgsea::fgsea() result tables (columns include pathway, pval, padj, NES) for each feature, ordered by p-value.

Details

The function works as follows:

  1. Hallmark gene sets are retrieved from MSigDB using msigdbr.

  2. A multivariate linear model is fitted for each gene using limma::lmFit.

  3. Empirical Bayes moderation is applied via limma::eBayes.

  4. For each feature:

    1. Differential expression results are extracted using topTable for the coefficient of that feature.

    2. Genes are ranked by moderated t-statistics.

    3. Hallmark GSEA is performed on the ranked gene list with fgsea::fgsea().

    4. Optionally, a dotplot of the top enriched pathways is generated.

Examples

if (FALSE) { # \dontrun{
  results <- compute_factor_gsea(RNA.tpm = expression_matrix,
                              features_df = feature_table,
                              plot_dot = TRUE,
                              top_n = 10,
                              file_name = "Feature1_vs_all")
} # }