Shared R helpers for Wu lab proteomics analysis.
Created 2026-08-04 while reorganizing OGlycoTM, which had normalize_intensity() /
filter_zero_na() copy-pasted into four scripts and run_limma_DE() into three. There was no
shared R code anywhere in the lab tree — all shared code was Python (mzml_utils,
spectrum_annotator_ddzby, the oglyco_* modules) — so the same TMT normalization and limma
pattern was being re-derived per project. NGlycoMM/data_analysis/limma_glycoprotein_diff.R is
another instance.
# once: git clone https://github.com/lfu46/wu-lab-r.git
source("path/to/wu-lab-r/wulab.R") # wulab.R sources everything under R/ relative to itself
intensity_cols <- c("Intensity.Tuni_1", "Intensity.Tuni_2", "Intensity.Tuni_3",
"Intensity.Ctrl_4", "Intensity.Ctrl_5", "Intensity.Ctrl_6")
norm <- quant |>
filter_zero_na(intensity_cols) |>
normalize_intensity(intensity_cols) # adds _sl and _sl_tmm columns
de <- run_limma_DE(
data = norm,
id_col = "Protein.ID",
metadata_cols = c("Entry.Name", "Gene", "Protein.Description"),
intensity_cols = paste0(intensity_cols, "_sl_tmm"), # limma runs on the normalized columns
design_matrix = Experiment_Model,
contrast_matrix = Contrast_Matrix
)It is a plain sourceable module, not a package — every consumer is an analysis script, and a
package would add a build/install step for no benefit. The R/ layout is package-shaped if that
ever changes.
| File | Provides |
|---|---|
R/tmt_normalize.R |
filter_zero_na(), normalize_intensity() — SL then edgeR TMM |
R/limma_de.R |
run_limma_DE(), summarize_DE(), classify_DE() |
R/theme_wulab.R |
theme_wulab(), save_figure(), wulab_palette, colors_cell, colors_glycan, colors_plddt |
Requires tidyverse, edgeR, limma, ggplot2.
tests/test_against_oglycotm.R re-runs normalization and limma against the published
OGlycoTM outputs. Byte-equivalent output is the bar before any caller is switched over.
$ Rscript tests/test_against_oglycotm.R
HEK293T rows 773 vs 773 max|diff| = 2.2e-16 PASS
HepG2 rows 673 vs 673 max|diff| = 2.22e-16 PASS
Jurkat rows 671 vs 671 max|diff| = 2.2e-16 PASS
limma matched 773/773 rows max|dlogFC| = 1.78e-14 max|dadjP| = 2.35e-13 PASS
R MODULE GATE: PASS
The test reads from the Expansion working copy if mounted, else the network share.
Both make the helpers safe to share; neither changes the numbers.
normalize_intensity()requiresintensity_cols. The original defaulted to OGlycoTM's six TMT channels; a shared helper must not assume an experimental design, so omitting it now errors instead of silently normalizing the wrong columns.classify_DE()takes anid_colargument. The original hard-codedProtein.ID, which made it unusable for site-level tables.
Worth lifting from OGlycoTM when a second consumer appears — noted so the candidates are not lost:
circlize_plot()/find_link_indices()(Figure4.R) — circular heatmap with cross-sector intralinks; already copied once intoNSequon_filter_comparison.R.write_protein_id_sheet()/write_de_sheet()/save_table()(fix_all_tables.R) — the Supporting-Information XLSX house style.wilcox_across_locations()/prepare_stat_annotations()(Figure5.R) — stacked significance brackets; the panel was cut from the paper but the helpers are the best in that repo.- A cached
enrichGO/gseGOwrapper (ac1c05438_si_004_GO_analysis.R,OGalNAc_GSEA.R) — caching toenrichment/*.csvis what lets figure scripts re-render in seconds instead of minutes.