Functional Analysis

Overview

This skill answers: What biological processes/pathways are represented by this gene set?

Two main approaches:

1. Over-representation Analysis (ORA): Test if genes are enriched in pathways
2. Gene Set Enrichment Analysis (GSEA): Use full ranked gene list

Plus specialized analysis: 3. CNV Analysis: Copy number variation in spatial context (tumor analysis)

Decision Tree: Which Method?

Q: What type of gene list do you have?
│
├─ Discrete gene list (DEGs, markers)
│   └─ Over-Representation Analysis (ORA)
│       ├─ Input: Gene list (up/down regulated)
│       ├─ Tests: Hypergeometric / Fisher's exact
│       └─ Output: Enriched pathways with p-values
│
├─ Ranked gene list (all genes with scores)
│   └─ Gene Set Enrichment Analysis (GSEA)
│       ├─ Input: All genes ranked by fold change/statistic
│       ├─ Tests: Running sum statistic
│       └─ Output: Enrichment scores, leading edge genes
│
└─ Spatial tumor data
    └─ CNV Analysis
        ├─ Infer copy number from expression
        ├─ Map CNV spatially
        └─ Identify subclones

Over-Representation Analysis (ORA)

When to Use

• You have a discrete gene list (e.g., significant DEGs)
• Want to know what pathways are enriched
• Quick, interpretable results needed

Databases

Database	Content	Best For
GO (Gene Ontology)	BP, MF, CC terms	Broad functional annotation
KEGG	Metabolic/signaling pathways	Pathway diagrams
Reactome	Curated pathways	Detailed mechanisms
MSigDB	Hallmarks, curated sets	Cancer, immunology
WikiPathways	Community pathways	Specific processes

Workflow

Step 1: Prepare Gene List

Requirements:

• Gene symbols or IDs (consistent naming)
• Typically 50-500 genes works best
• Separate up/down regulated if directional

Step 2: Run ORA

Use analyze_enrichment tool with:

• gene_list: Your genes of interest
• database: "GO_BP", "KEGG", "Reactome", etc.
• background: All detected genes (important!)

Step 3: Interpret Results

Key outputs:

• term: Pathway/GO term name
• pvalue / padj: Significance
• odds_ratio: Enrichment strength
• gene_count: Genes in overlap
• genes: Specific overlapping genes

Significance criteria:

Typically significant:
- padj < 0.05
- gene_count >= 3 (robust overlap)
- odds_ratio > 2 (meaningful enrichment)

Gene Set Enrichment Analysis (GSEA)

When to Use

• You have scores for all genes (not just significant ones)
• Want to detect subtle but coordinated changes
• Directional enrichment matters (up vs down)

Advantages over ORA

• Uses information from all genes
• Detects subtle coordinated changes
• No arbitrary threshold for "significant" genes
• Better statistical properties

Workflow

Step 1: Rank Genes

Rank all genes by:

• Log fold change (signed)
• -log10(p) × sign(logFC) (recommended)
• Test statistic

Step 2: Run GSEA

Use analyze_enrichment tool with:

• method: "gsea"
• ranking: Pre-computed gene ranking
• database: Gene set database

Step 3: Interpret Results

Key outputs:

• NES: Normalized Enrichment Score
  • Positive: Enriched in highly ranked genes
  • Negative: Enriched in lowly ranked genes
• pvalue / padj: Significance
• leading_edge: Core genes driving enrichment

Enrichment plot interpretation:

• Running enrichment score curve
• Peaks show where gene set is concentrated
• Leading edge = genes before peak

CNV Analysis (Tumor Spatial Data)

Purpose

Infer copy number variations from spatial transcriptomics for:

• Tumor heterogeneity mapping
• Clone identification
• Spatial organization of subclones

Methods

Method	Approach	Best For
inferCNV	Reference-based	scRNA-seq, validated
CopyKAT	Reference-free	No normal reference
SCEVAN	Fast, integrated	Large datasets

Workflow

Step 1: Define Reference

For CNV inference, need:

• Normal cells as reference (from annotation)
• Or use reference-free method

Step 2: Run CNV Analysis

Use analyze_cnv tool with:

• method: "infercnv", "copykat", or "scevan"
• reference_group: Normal cell label (if applicable)

Step 3: Interpret Results

CNV outputs:

• cnv_score: Per-cell CNV intensity
• clone_id: Inferred subclone assignment
• Chromosome-level gains/losses

Spatial visualization:

• Map CNV scores spatially
• Identify clone spatial organization
• Correlate with histology

Visualization

Enrichment Visualizations

Plot Type	Use Case
Bar plot	Top enriched terms
Dot plot	Significance + gene count
Network	Term relationships
GSEA plot	Enrichment curve
Heatmap	Gene-pathway matrix

Use visualize_data with plot_type="enrichment".

CNV Visualizations

Plot Type	Use Case
Heatmap	CNV across chromosomes
Spatial	CNV score on tissue
Clone map	Subclone distribution

Common Analysis Scenarios

Scenario 1: Understand DEG Function

Goal: What pathways are activated in disease?
1. Get DEG list from differential analysis
2. Split into up/down regulated
3. Run ORA on each separately
4. Compare enriched pathways

Scenario 2: Pathway-Level Changes

Goal: Are cancer hallmarks affected?
1. Rank all genes by fold change
2. Run GSEA with MSigDB Hallmarks
3. Identify significantly enriched hallmarks
4. Examine leading edge genes

Scenario 3: Tumor Clone Mapping

Goal: Map tumor heterogeneity spatially
1. Annotate normal cells (if present)
2. Run CNV inference
3. Cluster by CNV profile to identify clones
4. Map clones spatially
5. Correlate with histology

Scenario 4: Immune Signature Analysis

Goal: What immune processes are active where?
1. Define immune gene signatures
2. Score each spot for signatures
3. Map signature scores spatially
4. Correlate with cell composition

Best Practices

Gene List Quality

• Remove duplicates: Each gene counted once
• Consistent IDs: Use symbols or Ensembl, not mixed
• Appropriate background: All expressed genes, not all genes
• Reasonable size: 50-500 genes typically best for ORA

Multiple Testing

• Always use adjusted p-values
• Report number of tests performed
• Consider redundancy in GO terms

Interpretation Caution

• Enrichment ≠ causation
• Multiple pathways may overlap
• Validate key findings experimentally

Troubleshooting

Issue	Cause	Solution
No significant terms	Gene list too small/diverse	Use GSEA, check gene IDs
Too many terms	Loose threshold	Tighten padj, filter redundant
Unexpected results	Wrong background	Use detected genes as background
Missing expected pathway	Gene naming mismatch	Check ID format (symbol vs Ensembl)

Output Summary

Successful functional analysis provides:

• Enriched pathways: Ranked by significance
• Gene-pathway mapping: Which genes in which pathways
• Biological interpretation: What processes are affected
• CNV landscape: Tumor heterogeneity map (if applicable)
• Downstream hypotheses: Mechanisms to investigate