gsea-enrichment-analysis

Original source / Raw SKILL.md

---
name: gsea-enrichment-analysis
title: GSEA and Pathway Enrichment Analysis
description: Gene set enrichment analysis with correct geneset format handling. Critical guidance for loading pathway databases and running enrichment in OmicVerse.
---

# GSEA and Pathway Enrichment Analysis

## Overview
This skill covers gene set enrichment analysis (GSEA) and pathway enrichment workflows in OmicVerse. It provides critical guidance on the correct data formats and API usage patterns to avoid common errors.

## Critical API Reference - Geneset Format

### IMPORTANT: Use Dictionary Format, NOT File Path!

The `ov.bulk.geneset_enrichment()` function requires a **dictionary** of gene sets, NOT a file path string. You must first load the geneset file using `ov.utils.geneset_prepare()`.

**CORRECT usage:**
```python
# Step 1: Download pathway database (if not already available)
ov.utils.download_pathway_database()

# Step 2: Load geneset file into dictionary format - REQUIRED!
pathways_dict = ov.utils.geneset_prepare(
    'genesets/GO_Biological_Process_2021.txt',  # or .gmt file
    organism='Human'  # or 'Mouse'
)

# Step 3: Now run enrichment with the DICTIONARY
enr = ov.bulk.geneset_enrichment(
    gene_list=deg_genes,
    pathways_dict=pathways_dict,  # Pass the DICTIONARY, not file path!
    pvalue_type='auto',
    organism='Human'
)
```

**WRONG - DO NOT USE:**
```python
# WRONG! Don't pass file path directly to geneset_enrichment!
# enr = ov.bulk.geneset_enrichment(
#     gene_list=deg_genes,
#     pathways_dict='genesets/GO_Biological_Process_2021.gmt'  # ERROR! String path doesn't work!
# )

# WRONG! geneset_enrichment expects dict, not file path
# enr = ov.bulk.geneset_enrichment(
#     gene_list=deg_genes,
#     pathways_dict='GO_Biological_Process_2021'  # ERROR!
# )
```

### File Format Support

| File Extension | Load Method | Notes |
|---------------|-------------|-------|
| `.txt` | `ov.utils.geneset_prepare()` | OmicVerse format |
| `.gmt` | `ov.utils.geneset_prepare()` | Standard GMT format |
| `.json` | `json.load()` then convert | Custom handling needed |

### Complete Enrichment Workflow

```python
import omicverse as ov

# 1. Setup
ov.plot_set()

# 2. Ensure pathway database is available
ov.utils.download_pathway_database()

# 3. Load gene sets - ALWAYS use geneset_prepare first!
go_bp = ov.utils.geneset_prepare('genesets/GO_Biological_Process_2021.txt', organism='Human')
go_mf = ov.utils.geneset_prepare('genesets/GO_Molecular_Function_2021.txt', organism='Human')
kegg = ov.utils.geneset_prepare('genesets/KEGG_2021_Human.txt', organism='Human')

# 4. Prepare gene list (e.g., from DEG analysis)
# Assuming dds is a pyDEG object with results
deg_genes = dds.result.loc[dds.result['sig'] != 'normal'].index.tolist()

# 5. Run enrichment with dictionary
enr_go_bp = ov.bulk.geneset_enrichment(
    gene_list=deg_genes,
    pathways_dict=go_bp,  # Dictionary, NOT file path!
    pvalue_type='auto',
    organism='Human'
)

# 6. Visualize results
ov.bulk.geneset_plot(enr_go_bp, figsize=(6, 8), num=10)

# 7. For multiple databases, combine into dict
enr_dict = {
    'GO_BP': enr_go_bp,
    'GO_MF': enr_go_mf,
    'KEGG': enr_kegg
}
colors_dict = {
    'GO_BP': '#1f77b4',
    'GO_MF': '#ff7f0e',
    'KEGG': '#2ca02c'
}
ov.bulk.geneset_plot_multi(enr_dict, colors_dict, num=5)
```

## Common Errors and Solutions

### Error: "FileNotFoundError" or "pathways_dict is not a dict"
**Cause**: Passing file path string instead of dictionary to `geneset_enrichment()`
**Solution**: First load with `ov.utils.geneset_prepare()`, then pass the returned dictionary

### Error: "Missing file 'genesets/GO_Biological_Process_2021.gmt'"
**Cause**: Pathway database not downloaded
**Solution**: Run `ov.utils.download_pathway_database()` first

### Error: "No enriched pathways found"
**Cause**: Gene list doesn't overlap with pathway genes, or organism mismatch
**Solution**:
- Verify gene symbols match (human vs mouse capitalization)
- Check `organism` parameter matches your data
- Ensure gene list has sufficient genes (>10 recommended)

## Pathway Databases Available

After running `ov.utils.download_pathway_database()`:
- `GO_Biological_Process_2021.txt`
- `GO_Molecular_Function_2021.txt`
- `GO_Cellular_Component_2021.txt`
- `KEGG_2021_Human.txt`
- `KEGG_2021_Mouse.txt`
- `Reactome_2022.txt`
- `WikiPathway_2023_Human.txt`
- And many more...

## Best Practices

1. **Always load genesets first**: Never pass file paths directly to `geneset_enrichment()`
2. **Check gene format**: Ensure gene symbols match (CAPS for human, Title case for mouse)
3. **Download once**: Run `download_pathway_database()` once per environment
4. **Specify organism**: Always set `organism='Human'` or `organism='Mouse'`
5. **Use background genes**: For more accurate results, provide `background` parameter

## Examples
- "Run GO enrichment on my DEG results using the correct geneset_prepare workflow"
- "Perform KEGG pathway analysis on upregulated genes with proper dictionary format"
- "Compare GO BP, MF, and KEGG enrichment results using geneset_plot_multi"

## References
- Tutorial notebook: [`t_deg.ipynb`](../../omicverse_guide/docs/Tutorials-bulk/t_deg.ipynb) (enrichment section)
- Pathway download: `ov.utils.download_pathway_database()`
- Quick reference: [`reference.md`](reference.md)