proteomics-enrichment

proteomics-enrichment

When to use

The user has a CSV listing proteins of interest (e.g. the significant subset from proteomics-de, or the PTM-target list from proteomics-ptm) and wants over-representation enrichment via Fisher's exact test, with BH-adjusted FDR.

This is a demo-only enrichment. The pathway database is the hard-coded 8-pathway DEMO_PATHWAYS dict at prot_enrichment.py:40-49 (each pathway has 5 fixed members). There is NO CLI flag to load a real KEGG / Reactome / MSigDB library. For production proteomics enrichment, export your significant-protein list and call bulkrna-enrichment (which has real ORA + GSEA + ssGSEA backends with hosted libraries).

Inputs & Outputs

Inputs

• File types: .csv
• Accepts artifact proteomics.differential_results (csv)

Outputs

• tables/enrichment_results.csv
• report.md
• result.json

Flow

1. Load CSV (--input <proteins.csv>) or generate a demo (--demo).
2. Pick the gene-list column: protein_id if present, otherwise the first column (prot_enrichment.py:256).
3. For each pathway in DEMO_PATHWAYS (prot_enrichment.py:40-49), run Fisher's exact test (prot_enrichment.py:138-150); apply BH FDR adjustment (prot_enrichment.py:167).
4. Write tables/enrichment_results.csv (prot_enrichment.py:267) + report.md + result.json (:277).

Gotchas

• Pathway database is HARD-CODED 8 demo pathways. prot_enrichment.py:40-49 defines 8 pathways × 5 genes each (e.g. cell-cycle, apoptosis, TCA-cycle). There is no CLI for loading real databases. The n_pathways_tested = 8 in result.json (:271) is a constant, not a function of input. For real enrichment, route to bulkrna-enrichment.
• Method is Fisher's exact, not hypergeometric. Mathematically equivalent for over-representation, but the script and report (prot_enrichment.py:4, 138-150) consistently say "Fisher's". Hypergeometric is the same distribution but the "Fisher's exact test" naming is what shows in the report.
• Default background ≠ a real proteome size. prot_enrichment.py:126-128 sets background_size = max(len(gene_set | all_pathway_genes), len(gene_set) + 1) — for the demo's 8 pathways that's ~40 + n_input. Always pass --background-size N (e.g. 20000 for human, 8000 for your detected proteome) for real enrichment — the auto-default produces meaningless p-values on a real dataset.
• --species is RECORDED-ONLY. prot_enrichment.py:237-241 accepts --species but the value is never used to switch databases or filter pathways — it's logged into result.json for reproducibility only.
• Gene-list column auto-detection: protein_id first, else first column. prot_enrichment.py:256 uses gene_col = "protein_id" if "protein_id" in df.columns else df.columns[0]. If your CSV has multiple ID columns (gene, uniprot, symbol), only protein_id is preferred — pre-rename the column you want enriched.
• --input REQUIRED unless --demo. prot_enrichment.py:251 raises ValueError("--input required when not using --demo").