DiffexpWedgeR

DiffexpWedgeR is an R command-line pipeline that performs differential expression analysis from gene-level count tables using edgeR (QLF + TREAT), generates QC and visualization outputs, and runs GO enrichment using limma’s goana/topGO workflow. The pipeline is driven by a JSON configuration file that defines samples, factors, thresholds, organism settings, and output paths.

Requirements

R (recommended >= 4.1)

Packages:

• edgeR
• org.Hs.eg.db
• GO.db
• topGO
• argparse
• dplyr
• tibble
• pheatmap
• EnhancedVolcano
• ggplot2
• jsonlite

Input

Count files

Each sample must provide a tab-separated count table with:

• Header row
• First column as gene identifiers (used as row names)
• One numeric count column

Example format:

GeneID\tCounts TP53\t123 BRCA1\t45

JSON configuration

The pipeline reads a JSON file passed via --config. It expects these fields:

• project.p_cutoff
• project.lfc_cutoff
• project.interaction
• project.data_source
• project.output_dir
• project.organism.species_code
• project.samples.sample_name
• project.samples.countFile
• project.samples.factors (per-sample list of {name, levels})

data_source must be one of:

• ncbi (expects gene identifiers as SYMBOL)
• ensembl (expects ENSEMBL IDs; version suffixes like .12 are removed)

interaction controls the model formula for multifactor designs:

• "True" uses full interaction (*)
• otherwise additive (+)

Usage

Run the script:

Rscript diffexp_wedger.R --config config.json

Output

The pipeline writes the following into project.output_dir:

Tables:

• raw_counts_table.txt
• raw_filtered_counts_table.txt
• samples_table.txt

QC plots:

• MDPlots.jpg
• PCA_plot.jpg
• Dispersion_plot.jpg
• Fitted_mean_ql_dispersion_plot.jpg (single-factor)
• Fitted_mean_ql_dispersion_plot_single.jpg (multi-factor)
• Fitted_mean_ql_dispersion_plot_multi.jpg (multi-factor)

Per comparison / coefficient outputs:

• _result_qlf_test_w<lfc_cutoff>cutoff.txt
• _summary_qlf_test_w<lfc_cutoff>cutoff.txt
• _MDPlot_w<lfc_cutoff>cutoff.jpg
• _volcano_plot.jpg
• _GO_results.txt
• _heatmap_log2_transformed.jpg (single-factor and multi-factor where applicable)

For single-factor designs, all pairwise contrasts between group levels are tested. For multi-factor designs, each model coefficient (excluding intercept) is tested.

Notes on filtering and thresholds

• Genes are filtered using edgeR filterByExpr.
• Normalization uses TMM.
• Differential testing uses quasi-likelihood framework and TREAT with an effect-size threshold based on lfc_cutoff.
• Significance summaries use p_cutoff applied to TREAT results.

Organism and GO enrichment

GO enrichment is computed using goana with Entrez IDs obtained via org.Hs.eg.db. This implementation is currently configured for human annotation mapping; for non-human organisms, the annotation database and mapping strategy must be adapted accordingly.