# Prompt: Cross-Study Disease Gene Expression Meta-Analysis
## Objective
Identify consistently dysregulated genes across multiple independent studies of a specific disease to discover robust biomarkers or therapeutic targets.
## Background
The Gene Expression Atlas Knowledge Graph integrates 243 studies with 797 assays profiling 152,879 genes. This resource enables meta-analysis across independent experiments to identify genes with reproducible differential expression patterns. Genes showing consistent dysregulation across multiple studies represent more robust candidates for further investigation than single-study findings.
## Task Description
Perform a cross-study meta-analysis to identify genes that are consistently differentially expressed in [DISEASE OF INTEREST] across multiple independent studies in the Gene Expression Atlas Knowledge Graph.
### Specific Steps:
1. **Identify Relevant Studies**
- Query the knowledge graph to find all studies examining [DISEASE OF INTEREST]
- Filter studies by relevant experimental factors (e.g., specific tissue types, cell types, developmental stages)
- Document the number of studies and assays available for analysis
2. **Extract Differential Expression Data**
- For each relevant study/assay, retrieve genes showing differential expression
- Collect quantitative metrics including:
- Effect sizes (log2 fold changes)
- Statistical significance (p-values, adjusted p-values)
- Direction of change (up-regulated vs. down-regulated)
- Note the experimental context (test vs. reference groups)
3. **Cross-Study Consistency Analysis**
- Identify genes appearing in multiple studies
- Assess consistency of:
- Direction of differential expression (consistently up or down)
- Effect size magnitude
- Statistical significance across studies
- Rank genes by:
- Number of studies showing differential expression
- Consistency of direction
- Average effect size
4. **Biological Context Integration**
- Link identified genes to:
- Biological processes (GO terms)
- Molecular pathways
- Protein domains
- Known disease associations
- Examine anatomical and cellular contexts where genes show consistent patterns
5. **Generate Insights**
- Identify top candidate biomarker genes (high consistency, large effect sizes)
- Highlight novel candidates not previously associated with the disease
- Compare findings across different tissues/cell types if applicable
- Suggest potential mechanisms based on pathway connections
## Example Queries
### Find Studies on a Specific Disease
```sparql
PREFIX biolink: <https://w3id.org/biolink/vocab/>
PREFIX genelab: <https://spoke.ucsf.edu/genelab/>
SELECT DISTINCT ?study ?title ?pubmed
WHERE {
?study a biolink:Study ;
genelab:project_title ?title .
OPTIONAL { ?study genelab:pubmed_id ?pubmed . }
FILTER(CONTAINS(LCASE(?title), "your_disease_keyword"))
}
```
### Get Differential Expression Data for Studies
```sparql
PREFIX biolink: <https://w3id.org/biolink/vocab/>
PREFIX genelab: <https://spoke.ucsf.edu/genelab/>
PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
SELECT ?study_id ?gene ?gene_label ?effect_size ?p_value
WHERE {
?study a biolink:Study ;
genelab:study_id ?study_id ;
biolink:has_output ?assay .
?expr a biolink:GeneExpressionMixin ;
biolink:subject ?assay ;
biolink:object ?gene .
?gene rdfs:label ?gene_label .
OPTIONAL { ?expr genelab:effect_size ?effect_size . }
OPTIONAL { ?expr genelab:p_value ?p_value . }
FILTER(CONTAINS(LCASE(?study_title), "your_disease_keyword"))
}
```
## Success Criteria
The analysis should produce:
1. **Ranked Gene List**: Genes ordered by cross-study consistency and effect size
2. **Study Coverage Matrix**: Which genes appear in which studies with their expression patterns
3. **Biological Context**: Pathways and processes enriched in consistently dysregulated genes
4. **Novel Discoveries**: Genes not previously well-characterized in the disease context
5. **Validation Recommendations**: Suggestions for experimental validation based on reproducibility
## Example Diseases to Explore
- **Diabetic nephropathy**: Multiple studies examining kidney tissue in diabetes
- **Breast cancer**: Studies across different subtypes and stages
- **Inflammatory bowel disease**: Colon tissue samples from IBD patients
- **Alzheimer's disease**: Brain tissue from affected patients
- **Cardiovascular disease**: Heart and vascular tissue studies
## Expected Outputs
1. **Summary Report**:
- Number of studies analyzed
- Total genes with differential expression
- Top 20-50 genes with consistent patterns
- Statistical summaries of reproducibility
2. **Visualization Suggestions**:
- Heatmap of gene expression across studies
- Volcano plots for individual studies
- Network diagram of gene-pathway connections
- Venn diagrams of overlapping gene sets
3. **Biological Interpretation**:
- Functional enrichment analysis results
- Known vs. novel disease gene associations
- Tissue/cell-type specific patterns
- Potential therapeutic implications
## Tips for Success
- Start with diseases that have 3+ studies for meaningful meta-analysis
- Consider tissue specificity - compare studies from the same anatomical context
- Look for both highly significant genes AND genes with large effect sizes
- Don't ignore genes that are consistently changed but with modest effect sizes
- Consider experimental platform differences when interpreting results
- Use PubMed IDs to review original study designs and validate findings
## Next Steps After Analysis
1. Cross-reference findings with external databases (GeneCards, DisGeNET, OMIM)
2. Design validation experiments for top candidates
3. Explore therapeutic potential using drug-gene interaction databases
4. Consider single-cell or spatial transcriptomics follow-up for tissue-specific findings
5. Publish meta-analysis findings to contribute to the field
