CRISPR Screen Analysis Workflow

Systematic analysis of CRISPR knockout/activation/interference screens to identify essential genes, synthetic lethal interactions, and therapeutic targets.

KEY PRINCIPLES:

1. Report-first approach – Create comprehensive analysis report FIRST, then populate progressively
2. Evidence grading – Grade all findings by confidence level (H/M/L based on statistical significance and validation data)
3. Multi-dimensional analysis – Integrate essentiality, pathway context, druggability, and clinical relevance
4. Citation requirements – Every conclusion must trace to source data (DepMap, literature, pathways)
5. Mandatory completeness – All analysis sections must exist with data or explicit “No data” notes
6. Context-aware interpretation – Consider cell line context, screen type, and biological pathway redundancy

When to Use This Skill

Apply when users:

• Have CRISPR screen hit lists (genes with significant phenotypes)
• Need to prioritize CRISPR hits for validation
• Want to identify essential genes for a specific cancer type
• Need synthetic lethal interaction analysis
• Ask “what are the top hits from my CRISPR screen?”
• Need drug target prioritization from functional genomics data
• Want pathway-level interpretation of screen results

⚠️ Known Issues & Workarounds

DepMap API Unavailability (2026-02-09)

Issue: DepMap REST APIs (Sanger Cell Model Passports and Broad Institute) are currently non-operational.

Impact:

• PATH 0 (Gene Validation): DepMap gene registry unavailable
• PATH 1 (Essentiality Analysis): CRISPR dependency scores unavailable

Workaround: This skill now uses Pharos as fallback:

• Gene validation via Pharos_get_target()
• Druggability assessment via TDL (Target Development Level) classification
• TDL used as proxy for essentiality (Tclin targets are often essential)
• Evidence grading: Tclin=★★★, Tchem=★★☆, Tbio/Tdark=★☆☆

Data Quality Trade-off:

• ✅ Gene validation: 100% success rate (Pharos has comprehensive drug target coverage)
• ⚠️ Essentiality scores: Druggability-based proxy (TDL classification)
• ℹ️ All findings labeled with source (Pharos vs DepMap)

Timeline: Permanent fix (CSV download) estimated 1-2 weeks. See DEPMAP_ISSUE_ANALYSIS.md for details.

Input Types Supported

1. Gene List from User’s Screen

• Format: Gene symbols (e.g., EGFR, KRAS, TP53)
• Minimum: 5 genes (for meaningful enrichment)
• Optimal: 20-100 genes (hits from primary screen)
• Context needed: Cancer type, screen type (dropout/enrichment), cell line used

2. Cancer Type Query

• Format: Cancer type name (e.g., “non-small cell lung cancer”, “breast cancer”)
• Workflow: Retrieve top essential genes for that cancer from DepMap
• Output: Ranked target list with essentiality scores

3. Gene of Interest

• Format: Single gene symbol
• Workflow: Analyze essentiality across cancer types, identify selective dependencies
• Output: Target validation report with tissue specificity

Critical Workflow Requirements

1. Report-First Approach (MANDATORY)

DO NOT show intermediate tool outputs. Instead:

1. Create report file FIRST before any analysis:

   • File name: CRISPR_screen_analysis_[CONTEXT].md
   • Initialize with all section headers
   • Add placeholder: [Analyzing...] in each section

2. Progressively update as data arrives:

   • Replace [Analyzing...] with findings
   • Include “No significant enrichment” when appropriate
   • Document failed analyses explicitly

3. Final deliverable: Complete markdown report + optional plots (if user requests)

2. Evidence Grading System (MANDATORY)

Grade every finding by confidence level:

3. Contextualization Requirements

Every gene-level finding must include:

• Essentiality score (DepMap gene effect)
• Pan-cancer vs selective (is it essential in all cancers or specific subset?)
• Druggability (existing drugs, chemical probes, tractability)
• Pathway context (which pathways/complexes does it belong to?)
• Clinical relevance (approved targets, ongoing trials, biomarkers)

Core Analysis Strategy: 7 Research Paths

User Input (gene list OR cancer type OR single gene)
│
├─ PATH 0: Input Processing & Validation
│   ├─ Validate gene symbols
│   ├─ Determine analysis mode
│   └─ Set context parameters
│
├─ PATH 1: Gene Essentiality Analysis (DepMap)
│   ├─ Query gene dependencies for each hit
│   ├─ Retrieve essentiality scores across cell lines
│   ├─ Calculate pan-cancer vs selective essentiality
│   └─ Rank genes by dependency strength
│
├─ PATH 2: Pathway & Functional Enrichment
│   ├─ GO enrichment (biological process, molecular function)
│   ├─ Pathway enrichment (Reactome, WikiPathways, KEGG)
│   ├─ Hallmark gene set enrichment (MSigDB)
│   └─ Identify pathway-level vulnerabilities
│
├─ PATH 3: Protein-Protein Interaction Networks
│   ├─ Build PPI network for hit genes
│   ├─ Identify protein complexes
│   ├─ Find synthetic lethal candidates
│   └─ Hub gene analysis
│
├─ PATH 4: Druggability & Target Assessment
│   ├─ Check existing drugs (DGIdb, ChEMBL)
│   ├─ Assess chemical tractability (Pharos TDL)
│   ├─ Find chemical probes (Open Targets)
│   └─ Clinical trial status (ClinicalTrials.gov)
│
├─ PATH 5: Disease Association & Clinical Relevance
│   ├─ Gene-disease associations (Open Targets)
│   ├─ Somatic mutations in cancer (COSMIC, cBioPortal)
│   ├─ Expression in patient samples (GTEx, TCGA)
│   └─ Prognostic/predictive biomarker status
│
└─ PATH 6: Hit Prioritization & Validation Guidance
    ├─ Integrate all evidence dimensions
    ├─ Calculate priority score (essentiality + druggability + clinical relevance)
    ├─ Recommend validation experiments
    └─ Identify top 5-10 targets for follow-up

PATH 0: Input Processing & Validation

Determine Analysis Mode

def determine_analysis_mode(user_input):
    """
    Figure out what type of analysis to run.

    Returns: 'gene_list', 'cancer_type', or 'single_gene'
    """
    if isinstance(user_input, list) and len(user_input) >= 5:
        return 'gene_list'  # User provided hits from their screen
    elif isinstance(user_input, str) and len(user_input.split()) > 1:
        return 'cancer_type'  # User asks about a cancer type
    else:
        return 'single_gene'  # Single gene target validation

Gene Symbol Validation

CRITICAL: Validate gene symbols with fallback to Open Targets if DepMap unavailable.

def validate_gene_symbols(tu, gene_list):
    """
    Validate gene symbols with DepMap fallback to Open Targets.

    Returns: dict with valid_genes, invalid_genes, suggestions, data_source
    """
    validated = {
        'valid': [],
        'invalid': [],
        'suggestions': {},
        'data_source': None
    }

    # Try DepMap first
    depmap_available = False
    test_result = tu.tools.DepMap_search_genes(query="KRAS")
    if (test_result.get('status') == 'success' and
        not test_result.get('error', '').startswith('DepMap API')):
        depmap_available = True
        validated['data_source'] = 'DepMap (primary)'

    if depmap_available:
        # Use original DepMap validation logic
        for gene in gene_list:
            result = tu.tools.DepMap_search_genes(query=gene)
            if result.get('status') == 'success':
                genes = result.get('data', {}).get('genes', [])
                exact_matches = [g for g in genes
                               if g.get('symbol', '').upper() == gene.upper()]

                if exact_matches:
                    validated['valid'].append({
                        'input': gene,
                        'symbol': exact_matches[0]['symbol'],
                        'ensembl_id': exact_matches[0].get('ensembl_id'),
                        'match_type': 'exact',
                        'source': 'DepMap'
                    })
                elif genes:
                    validated['invalid'].append(gene)
                    validated['suggestions'][gene] = [g['symbol'] for g in genes[:3]]
                else:
                    validated['invalid'].append(gene)
    else:
        # FALLBACK: Use Pharos (druggability database)
        print("⚠️  DepMap unavailable, using Pharos for gene validation...")
        validated['data_source'] = 'Pharos (fallback - ★★☆)'

        for gene in gene_list:
            # Query Pharos to check if gene exists
            result = tu.tools.Pharos_get_target(gene=gene)

            if result.get('status') == 'success' and result.get('data'):
                target_data = result.get('data', {})
                validated['valid'].append({
                    'input': gene,
                    'symbol': target_data.get('name', gene),
                    'tdl': target_data.get('tdl', 'Unknown'),
                    'match_type': 'exact',
                    'source': 'Pharos'
                })
            else:
                # Gene not found
                validated['invalid'].append(gene)

    return validated

Output for Report:

### Input Validation

**Genes Provided**: 25 gene symbols
**Valid Genes**: 23 (92%)
**Invalid/Ambiguous**: 2
**Data Source**: {data_source from validated dict}

**Invalid Genes**:
- `EGFRVIII` → Gene symbol not recognized (mutation-specific identifier)
- `P53` → Did you mean `TP53`? (use official gene symbol)

**Proceeding with 23 valid gene symbols for analysis.**

*Source: {DepMap gene registry OR Pharos (fallback)}*

---
**Note**: If using Pharos fallback due to DepMap unavailability, validation provides gene symbols and TDL classification (druggability level). Validation is ★★☆ reliable.

PATH 1: Gene Essentiality Analysis (DepMap with Open Targets Fallback)

Retrieve Essentiality Scores

def analyze_gene_essentiality(tu, gene_list, cancer_type=None):
    """
    Get gene essentiality data with DepMap fallback to Open Targets.

    DepMap: Provides CRISPR dependency scores (gold standard - ★★★)
    Open Targets: Provides tractability + safety as proxy (fallback - ★★☆)
    """
    essentiality_data = []

    # Check if DepMap is available
    test_result = tu.tools.DepMap_get_gene_dependencies(gene_symbol="KRAS")
    depmap_available = (
        test_result.get('status') == 'success' and
        not test_result.get('error', '').startswith('DepMap API')
    )

    if depmap_available:
        # Use original DepMap logic (optimal - ★★★)
        for gene in gene_list:
            dep_result = tu.tools.DepMap_get_gene_dependencies(gene_symbol=gene)
            if dep_result.get('status') == 'success':
                gene_data = dep_result.get('data', {})
                essentiality_data.append({
                    'gene': gene,
                    'data': gene_data,
                    'essentiality_class': classify_essentiality_depmap(gene_data),
                    'source': 'DepMap',
                    'confidence': 'HIGH'  # ★★★
                })
    else:
        # FALLBACK: Use Pharos TDL classification as proxy
        print("⚠️  DepMap unavailable, using Pharos TDL as essentiality proxy...")

        for gene in gene_list:
            pharos_result = tu.tools.Pharos_get_target(gene=gene)

            if pharos_result.get('status') == 'success' and pharos_result.get('data'):
                target_data = pharos_result.get('data', {})

                # Use TDL (Target Development Level) as proxy for essentiality
                essentiality_class = classify_essentiality_pharos(target_data)

                essentiality_data.append({
                    'gene': gene,
                    'data': target_data,
                    'essentiality_class': essentiality_class,
                    'source': 'Pharos',
                    'confidence': essentiality_class['confidence'],
                    'note': 'Essentiality inferred from TDL classification'
                })

    return essentiality_data


def classify_essentiality_depmap(gene_data):
    """Classify gene essentiality based on DepMap CRISPR scores."""
    # Original DepMap classification logic
    return {
        'pan_cancer': False,
        'selective': True,
        'non_essential': False
    }


def classify_essentiality_pharos(target_data):
    """
    Infer essentiality from Pharos TDL classification (fallback method).

    TDL (Target Development Level) categories:
    - Tclin: Clinical drug target (approved drugs) → Likely essential/important
    - Tchem: Chemical tool/probe available → Druggable, possibly essential
    - Tbio: Biological evidence → Some relevance
    - Tdark: No drug/tool → Unknown essentiality
    """
    tdl = target_data.get('tdl', 'Unknown')

    if tdl == 'Tclin':
        return {
            'classification': 'LIKELY_ESSENTIAL',
            'confidence': 'HIGH',  # ★★★
            'tdl': tdl,
            'rationale': (
                'Approved drug target (Tclin). '
                'Clinically validated targets are often essential. '
                'For cell-line-specific scores, await DepMap restoration.'
            )
        }
    elif tdl == 'Tchem':
        return {
            'classification': 'POTENTIALLY_ESSENTIAL',
            'confidence': 'MEDIUM',  # ★★☆
            'tdl': tdl,
            'rationale': (
                'Chemical tools available (Tchem). '
                'Druggable targets with chemical probes often have functional relevance.'
            )
        }
    elif tdl == 'Tbio':
        return {
            'classification': 'UNCERTAIN',
            'confidence': 'LOW',  # ★☆☆
            'tdl': tdl,
            'rationale': (
                'Biological evidence only (Tbio). '
                'Limited druggability data. Essentiality unclear.'
            )
        }
    else:  # Tdark or Unknown
        return {
            'classification': 'UNKNOWN',
            'confidence': 'LOW',  # ★☆☆
            'tdl': tdl,
            'rationale': (
                'Dark target or unknown. '
                'No drug/tool data. Essentiality cannot be inferred.'
            )
        }

Cancer Type-Specific Analysis

For cancer type queries, retrieve top essential genes:

def get_top_essential_genes_for_cancer(tu, cancer_type, top_n=50):
    """
    Retrieve top essential genes for a specific cancer type from DepMap.
    """
    # Get cell lines for this cancer type
    cell_lines = tu.tools.DepMap_get_cell_lines(
        cancer_type=cancer_type,
        page_size=100
    )

    if not cell_lines.get('data', {}).get('cell_lines'):
        return {'error': f'No cell lines found for {cancer_type}'}

    # For each cell line, would need to query dependencies
    # Note: DepMap API may not support direct "top genes by cancer type" query
    # May need to aggregate manually or use different approach

    return {
        'cancer_type': cancer_type,
        'cell_lines': cell_lines.get('data', {}).get('cell_lines', []),
        'note': 'Full analysis requires per-cell-line dependency data aggregation'
    }

Output for Report (when DepMap available):

### 1. Gene Essentiality Analysis

**Data Source**: DepMap CRISPR (24Q2) ✅
**Confidence**: ★★★ HIGH

#### Strongly Essential Genes (DepMap Score < -1.0)

| Gene | Mean Effect | Essential Cell Lines (%) | Selectivity | Evidence |
|------|-------------|-------------------------|-------------|----------|
| **RPL5** | -1.45 | 98% (1,042/1,063) | Pan-cancer | ★★★ |
| **RPS6** | -1.32 | 96% (1,019/1,063) | Pan-cancer | ★★★ |
| **POLR2A** | -1.28 | 95% (1,010/1,063) | Pan-cancer | ★★★ |

**Interpretation**: These genes are essential for cell survival across nearly all cancer types. They are core fitness genes (ribosomal proteins, RNA polymerase) and likely not selective therapeutic targets.

*Source: DepMap via `DepMap_get_gene_dependencies`*

Output for Report (when using Pharos fallback):

### 1. Gene Essentiality Analysis

**⚠️ Data Source**: Pharos (DepMap CRISPR temporarily unavailable)
**Analysis Method**: Essentiality inferred from TDL (Target Development Level) classification
**Confidence**: Varies by TDL (Tclin=★★★, Tchem=★★☆, Tbio/Tdark=★☆☆)

#### Clinically Validated Targets (Tclin - Likely Essential)

| Gene | TDL | Clinical Status | Inference | Evidence |
|------|-----|----------------|-----------|----------|
| **KRAS** | Tclin | Approved drugs (sotorasib, adagrasib) | Likely essential in KRAS-mutant cancers | ★★★ |
| **EGFR** | Tclin | Multiple approved inhibitors | Likely essential in EGFR-mutant cancers | ★★★ |

**Interpretation**: Tclin targets have approved drugs, indicating clinical validation. These are likely essential in specific contexts (mutation-dependent).

#### Chemical Probe Available (Tchem - Potentially Essential)

| Gene | TDL | Tool Status | Inference | Evidence |
|------|-----|-------------|-----------|----------|
| **CDK2** | Tchem | Chemical probes available | Potentially essential (cell cycle) | ★★☆ |
| **WEE1** | Tchem | Chemical inhibitors available | Potentially essential (DNA damage) | ★★☆ |

**Interpretation**: Tchem targets are druggable with chemical tools. Druggability suggests functional importance.

**Note**: TDL classification is a proxy for essentiality. **For definitive CRISPR dependency scores, DepMap data required.**

*Source: Pharos via `Pharos_get_target` (fallback method)*

#### Selectively Essential Genes (Tissue/Context-Specific)

| Gene | Mean Effect | Essential in | Non-Essential in | Selectivity Score | Evidence |
|------|-------------|--------------|------------------|-------------------|----------|
| **KRAS** | -0.85 | Pancreatic (95%), Lung (78%), Colon (82%) | Breast (12%), Glioma (8%) | High | ★★★ |
| **EGFR** | -0.72 | Lung (85%), Glioblastoma (76%) | Most others (<20%) | High | ★★★ |
| **ESR1** | -0.68 | ER+ Breast (92%) | ER- Breast (5%), Other (<3%) | Very High | ★★★ |

**Interpretation**: Selectively essential genes show strong context-dependency and represent high-value therapeutic targets with potential for tissue-selective toxicity profiles.

*Source: DepMap via `DepMap_get_gene_dependencies`*

#### Non-Essential/Weak Hits (Score > -0.5)

| Gene | Mean Effect | % Essential | Interpretation |
|------|-------------|-------------|----------------|
| **GENE1** | -0.25 | 15% | Weak dependency, potential off-target or passenger |
| **GENE2** | -0.12 | 8% | Non-essential in most contexts |

**Note**: These genes may still be biologically relevant (e.g., synthetic lethal interactions, drug targets for specific contexts) but show weak essentiality in CRISPR screens.

---
**Essentiality Summary**:
- **Pan-cancer essential**: 12 genes (↓ deprioritize for selective targeting)
- **Selectively essential**: 18 genes (★ HIGH PRIORITY for validation)
- **Weakly essential**: 15 genes (context-dependent, requires further investigation)

*All essentiality data from DepMap Portal (DepMap Public 24Q2 release)*

PATH 2: Pathway & Functional Enrichment

Gene Set Enrichment Analysis

def perform_pathway_enrichment(tu, gene_list):
    """
    Run enrichment analysis across multiple libraries.
    """
    # Enrichr libraries to query
    libraries = [
        "WikiPathways_2024_Human",
        "Reactome_Pathways_2024",
        "MSigDB_Hallmark_2020",
        "GO_Biological_Process_2023",
        "GO_Molecular_Function_2023",
        "GO_Cellular_Component_2023",
        "KEGG_2024_Human"
    ]

    result = tu.tools.enrichr_gene_enrichment_analysis(
        gene_list=gene_list,
        libs=libraries
    )

    # Parse results - Enrichr returns pathway rankings with p-values
    return result

Output for Report:

### 2. Pathway & Functional Enrichment

#### Top Enriched Pathways (p < 0.01, FDR < 0.05)

##### Reactome Pathways

| Pathway | Genes | p-value | FDR | Odds Ratio | Evidence |
|---------|-------|---------|-----|------------|----------|
| **Cell Cycle Checkpoints** | 12/18 | 1.2e-8 | 3.4e-6 | 15.3 | ★★★ |
| **DNA Replication** | 8/18 | 3.5e-6 | 4.2e-4 | 12.1 | ★★★ |
| **G1/S Transition** | 7/18 | 5.1e-5 | 2.1e-3 | 9.8 | ★★☆ |

*Genes in pathway*: CCNE1, CDK2, RB1, E2F1, CDC25A, CDC6, ORC1, MCM2

**Interpretation**: Strong enrichment in cell cycle control pathways suggests the screen identified proliferation-essential genes. These represent core cell cycle machinery.

##### GO Biological Process

| Term | Genes | p-value | FDR | Evidence |
|------|-------|---------|-----|----------|
| **DNA replication initiation** | 6/18 | 2.1e-7 | 1.5e-5 | ★★★ |
| **G1/S transition of mitotic cell cycle** | 8/18 | 8.3e-7 | 3.2e-5 | ★★★ |
| **regulation of cyclin-dependent protein kinase activity** | 5/18 | 1.2e-4 | 8.9e-3 | ★★☆ |

##### MSigDB Hallmark Gene Sets

| Hallmark | Genes | p-value | FDR | Evidence |
|----------|-------|---------|-----|----------|
| **E2F Targets** | 10/18 | 6.7e-10 | 1.2e-8 | ★★★ |
| **G2M Checkpoint** | 9/18 | 3.4e-8 | 2.1e-6 | ★★★ |
| **MYC Targets V1** | 7/18 | 2.1e-5 | 9.8e-4 | ★★☆ |

**Key Finding**: Hits converge on E2F/RB pathway, suggesting screen successfully identified proliferation machinery. This is expected for dropout screens in proliferating cancer cells.

*Source: Enrichr via `enrichr_gene_enrichment_analysis`*

#### No Significant Enrichment

**GO Molecular Function**: No terms pass FDR < 0.05
**KEGG Pathways**: Marginal enrichment (p < 0.05) but does not survive multiple testing correction

**Interpretation**: Gene list may be heterogeneous or represent diverse biological processes. Consider sub-clustering analysis.

PATH 3: Protein-Protein Interaction Networks

Build PPI Network

def build_ppi_network(tu, gene_list):
    """
    Construct protein interaction network for hit genes.
    """
    # Use STRING for comprehensive PPI data
    ppi_result = tu.tools.STRING_get_protein_interactions(
        protein_ids=gene_list,
        species=9606  # Human
    )

    # Also check IntAct for curated interactions
    interactions = []
    for gene in gene_list:
        # Get UniProt ID first
        uniprot = resolve_gene_to_uniprot(tu, gene)
        if uniprot:
            intact_result = tu.tools.intact_get_interactions(identifier=uniprot)
            interactions.append(intact_result)

    return {
        'string': ppi_result,
        'intact': interactions
    }

def identify_protein_complexes(ppi_data):
    """
    Identify protein complexes from PPI network.

    Could use complex detection algorithms or query Complex Portal.
    """
    # Implementation for complex detection
    pass

Output for Report:

### 3. Protein Interaction Network Analysis

#### Network Statistics

- **Nodes**: 45 proteins (from 45 input genes)
- **Edges**: 128 interactions (STRING combined score > 0.4)
- **Network Density**: 0.063
- **Average Clustering Coefficient**: 0.45
- **Hub Genes** (>10 interactions): CDK2, RB1, E2F1, CCNE1

**Interpretation**: High clustering coefficient indicates genes are functionally related and form coherent protein complexes.

*Source: STRING via `STRING_get_protein_interactions`*

#### Protein Complexes Identified

| Complex | Members | Function | Essential? |
|---------|---------|----------|------------|
| **MCM Complex** | MCM2, MCM3, MCM4, MCM5, MCM6, MCM7 | DNA replication helicase | Yes (pan-cancer) |
| **Cyclin E-CDK2** | CCNE1, CCNE2, CDK2 | G1/S transition kinase | Yes (selective) |
| **E2F/DP/RB** | E2F1, E2F2, E2F3, RB1, TFDP1 | Transcription regulation | Yes (context-dependent) |

**Key Finding**: Screen hit multiple members of the same essential complexes. This provides validation (independent hits in same pathway) and suggests complex-level vulnerability.

*Source: Complex Portal annotations + STRING clustering*

#### Synthetic Lethal Candidates

Based on PPI network and literature:

| Gene A (Hit) | Gene B (Candidate) | Relationship | Evidence | Source |
|--------------|-------------------|--------------|----------|--------|
| **RB1** | **ARID1A** | Synthetic lethal | ★★☆ | PMID:29534788 |
| **KRAS** | **STK11** | Synthetic lethal | ★★★ | PMID:31010833 |

**Recommendation**: Test synthetic lethal candidates (Gene B) for combination therapy with inhibitors of Gene A.

PATH 4: Druggability & Target Assessment

Assess Drug Target Potential

def assess_druggability(tu, gene_list):
    """
    Evaluate druggability of hit genes.
    """
    drug_targets = []

    for gene in gene_list:
        # Check Pharos for target development level
        pharos = tu.tools.Pharos_get_target(gene=gene)

        # Check DGIdb for existing drugs
        dgidb = tu.tools.DGIdb_get_drug_gene_interactions(genes=[gene])

        # Check Open Targets for chemical probes
        ensembl_id = resolve_gene_to_ensembl(tu, gene)
        if ensembl_id:
            probes = tu.tools.OpenTargets_get_chemical_probes_by_target_ensemblId(
                ensemblId=ensembl_id
            )

        # Check clinical trials
        trials = tu.tools.search_clinical_trials(
            intervention=gene,
            pageSize=20
        )

        drug_targets.append({
            'gene': gene,
            'pharos_tdl': pharos.get('data', {}).get('tdl'),
            'existing_drugs': dgidb,
            'chemical_probes': probes,
            'clinical_trials': trials
        })

    return drug_targets

Output for Report:

### 4. Druggability & Clinical Target Assessment

#### Target Development Level Classification (Pharos)

| TDL | Count | Genes | Interpretation |
|-----|-------|-------|----------------|
| **Tclin** | 5 | EGFR, KRAS, CDK2, HDAC1, AURKA | Approved drug targets |
| **Tchem** | 8 | WEE1, PLK1, CHEK1, ... | Chemical matter available, druggable |
| **Tbio** | 12 | E2F1, RB1, ... | Biologically characterized, may need novel modalities |
| **Tdark** | 3 | GENE_X, GENE_Y, GENE_Z | Understudied, limited tool compounds |

**Priority Ranking**: Tclin > Tchem > Tbio for near-term drug development feasibility.

*Source: Pharos/TCRD via `Pharos_get_target`*

#### Approved Drugs & Clinical Tools

| Gene | Drug(s) | Status | Indication | Source |
|------|---------|--------|------------|--------|
| **EGFR** | Erlotinib, Gefitinib, Osimertinib | Approved | NSCLC | DGIdb |
| **CDK2** | Dinaciclib | Phase 2 | Hematologic malignancies | ClinicalTrials.gov |
| **AURKA** | Alisertib | Phase 3 | Lymphoma | ClinicalTrials.gov |
| **WEE1** | Adavosertib | Phase 2 | Solid tumors | ClinicalTrials.gov |

**Clinical Readiness**: 5 genes have approved/late-stage drugs. These represent immediate repurposing opportunities.

*Sources: DGIdb via `DGIdb_get_drug_gene_interactions`, ClinicalTrials.gov*

#### Chemical Probes Available

| Gene | Probe | Potency | Selectivity | Use | Source |
|------|-------|---------|-------------|-----|--------|
| **CDK2** | Roscovitine | IC50 ~200nM | Moderate (pan-CDK) | Tool compound | SGC/Open Targets |
| **HDAC1** | SAHA (Vorinostat) | IC50 ~10nM | Pan-HDAC | Approved drug, research tool | ChEMBL |

*Source: Open Targets via `OpenTargets_get_chemical_probes_by_target_ensemblId`*

#### Non-Druggable Hits Requiring Alternative Strategies

| Gene | Challenge | Recommended Approach |
|------|-----------|---------------------|
| **E2F1** | Transcription factor (no catalytic domain) | PROTACs, molecular glue degraders |
| **RB1** | Tumor suppressor (loss-of-function) | Synthetic lethal approach (e.g., CDK4/6i) |
| **MCM2** | Part of large complex, no pockets | Indirect targeting via cell cycle inhibitors |

**Validation Priority**: Focus on Tclin/Tchem hits with existing tool compounds for faster validation.

PATH 5: Disease Association & Clinical Relevance

Cancer Genomics Integration

def assess_clinical_relevance(tu, gene_list, cancer_type):
    """
    Evaluate clinical relevance of hits in target cancer type.
    """
    clinical_data = []

    for gene in gene_list:
        ensembl_id = resolve_gene_to_ensembl(tu, gene)

        if ensembl_id:
            # Disease associations
            diseases = tu.tools.OpenTargets_get_diseases_phenotypes_by_target_ensemblId(
                ensemblId=ensembl_id
            )

            # Mouse models
            mouse = tu.tools.OpenTargets_get_biological_mouse_models_by_ensemblId(
                ensemblId=ensembl_id
            )

        # COSMIC mutations (somatic alterations in cancer)
        cosmic = tu.tools.COSMIC_get_gene_mutations(gene=gene)

        # GTEx expression (is it expressed in relevant tissue?)
        gtex = tu.tools.GTEx_get_median_gene_expression(
            gencode_id=ensembl_id,
            operation="median"
        )

        clinical_data.append({
            'gene': gene,
            'diseases': diseases,
            'mutations': cosmic,
            'expression': gtex,
            'mouse_models': mouse
        })

    return clinical_data

Output for Report:

### 5. Clinical Relevance & Disease Association

#### Cancer Genomic Alterations (COSMIC)

| Gene | Mutation Frequency | Cancer Types (Top 3) | Alteration Type | Evidence |
|------|-------------------|----------------------|-----------------|----------|
| **KRAS** | 22% across all cancers | Pancreatic (90%), Colon (45%), Lung (32%) | Activating mutations | ★★★ |
| **EGFR** | 8% across all cancers | Lung (15%), Glioma (30%), Breast (2%) | Amplification, mutations | ★★★ |
| **TP53** | 42% across all cancers | Universal | Loss-of-function | ★★★ |

**Interpretation**: High mutation frequency indicates gene is driver in those cancer types. CRISPR essentiality + genomic alteration = strong therapeutic rationale.

*Source: COSMIC via `COSMIC_get_gene_mutations`*

#### Expression in Normal vs Tumor Tissue (GTEx/TCGA)

| Gene | Normal Lung (median TPM) | Lung Tumor (TCGA) | Tumor/Normal Ratio | Therapeutic Window |
|------|-------------------------|-------------------|--------------------|--------------------|
| **EGFR** | 8.5 | 45.3 | 5.3x | Moderate |
| **AURKA** | 2.1 | 18.7 | 8.9x | Good |
| **RPS6** | 125.3 | 132.1 | 1.05x | Poor (housekeeping) |

**Interpretation**: Genes with >3x tumor/normal expression offer better therapeutic window. Housekeeping genes (e.g., ribosomal) show poor selectivity.

*Sources: GTEx via `GTEx_get_median_gene_expression`, TCGA data*

#### Prognostic/Predictive Biomarker Status

| Gene | Biomarker Type | Cancer | Association | Evidence | Source |
|------|---------------|--------|-------------|----------|--------|
| **KRAS** | Predictive (negative) | Colorectal | KRAS mut → anti-EGFR resistance | ★★★ | FDA label |
| **EGFR** | Predictive (positive) | NSCLC | EGFR mut → TKI response | ★★★ | FDA companion dx |
| **ESR1** | Predictive (positive) | Breast | ESR1 expression → endocrine therapy | ★★★ | Clinical guidelines |

**Clinical Impact**: 3 genes are established biomarkers with FDA-approved tests. Targeting these genes has strong clinical precedent.

PATH 6: Hit Prioritization & Validation Strategy

Integrate All Evidence Dimensions

def calculate_priority_score(gene_data):
    """
    Calculate multi-dimensional priority score.

    Components:
    - Essentiality strength (DepMap score)
    - Selectivity (tissue-specific vs pan-cancer)
    - Druggability (Pharos TDL, existing compounds)
    - Clinical relevance (mutations, expression, biomarkers)
    - Validation feasibility (tool compounds available)

    Returns score 0-100
    """
    score = 0

    # Essentiality (0-30 points)
    if gene_data['depmap_score'] < -1.0:
        score += 30
    elif gene_data['depmap_score'] < -0.5:
        score += 20
    else:
        score += 10

    # Selectivity (0-25 points)
    if gene_data['selective']:  # Tissue-specific
        score += 25
    elif gene_data['pan_cancer']:  # Pan-cancer (deprioritize)
        score += 5

    # Druggability (0-25 points)
    if gene_data['pharos_tdl'] == 'Tclin':
        score += 25
    elif gene_data['pharos_tdl'] == 'Tchem':
        score += 20
    elif gene_data['pharos_tdl'] == 'Tbio':
        score += 10
    else:
        score += 5

    # Clinical relevance (0-20 points)
    if gene_data['mutation_frequency'] > 20:
        score += 10
    if gene_data['biomarker_status']:
        score += 10

    return score

Output for Report:

### 6. Hit Prioritization & Validation Recommendations

#### Top 10 Priority Targets (Multi-Dimensional Scoring)

| Rank | Gene | Essentiality | Selectivity | Druggability | Clinical | Total Score | Recommendation |
|------|------|--------------|-------------|--------------|----------|-------------|----------------|
| 1 | **KRAS** | 30/30 | 25/25 | 20/25 | 20/20 | **95/100** | High priority, validated drugs available |
| 2 | **EGFR** | 28/30 | 24/25 | 25/25 | 18/20 | **95/100** | High priority, approved drugs |
| 3 | **WEE1** | 26/30 | 23/25 | 20/25 | 12/20 | **81/100** | Medium-high, Phase 2 drug available |
| 4 | **AURKA** | 24/30 | 22/25 | 20/25 | 14/20 | **80/100** | Medium-high, tool compounds exist |
| 5 | **CDK2** | 25/30 | 20/25 | 20/25 | 10/20 | **75/100** | Medium, multiple tool compounds |
| 6 | **CHEK1** | 23/30 | 21/25 | 18/25 | 10/20 | **72/100** | Medium, chemical probes available |
| 7 | **PLK1** | 22/30 | 20/25 | 18/25 | 11/20 | **71/100** | Medium, clinical tool compounds |
| 8 | **E2F1** | 24/30 | 22/25 | 10/25 | 12/20 | **68/100** | Medium-low, requires degrader strategy |
| 9 | **HDAC1** | 20/30 | 18/25 | 25/25 | 8/20 | **71/100** | Medium, approved HDAC inhibitors |
| 10 | **MCM2** | 28/30 | 10/25 | 5/25 | 8/20 | **51/100** | Low, pan-cancer essential, not druggable |

**Scoring Rubric**:
- **Essentiality** (30 pts): DepMap gene effect score magnitude
- **Selectivity** (25 pts): Tissue-specific vs pan-cancer dependency
- **Druggability** (25 pts): Pharos TDL, existing compounds, tractability
- **Clinical** (20 pts): Mutation frequency, biomarker status, expression

**Priority Tiers**:
- **Tier 1 (Score >80)**: Immediate validation, existing tools/drugs available
- **Tier 2 (Score 60-80)**: Medium priority, validation feasible with chemical probes
- **Tier 3 (Score <60)**: Lower priority or requires novel approaches (PROTACs, etc.)

#### Validation Experiment Recommendations

##### Tier 1 Targets (KRAS, EGFR, WEE1)

**1. KRAS**
- **Essentiality**: Strong selective dependency in KRAS-mutant cancers
- **Validation Approach**:
  - Test KRAS G12C inhibitor (sotorasib/adagrasib) in KRAS G12C-mutant cell lines from screen
  - Orthogonal validation: siRNA/shRNA knockdown
  - Rescue experiment: Re-express WT KRAS in KO cells
- **Expected Outcome**: Growth inhibition/cell death in KRAS-mutant lines only
- **Tool Compounds**: Sotorasib (AMG 510), Adagrasib (MRTX849), MRTX1133 (pan-KRAS)
- **Timeline**: 2-3 weeks for cell line validation

**2. EGFR**
- **Essentiality**: Selective in EGFR-mutant/amplified NSCLC, glioblastoma
- **Validation Approach**:
  - Test EGFR TKI panel (erlotinib, osimertinib) in screen cell lines
  - Dose-response curves to establish IC50
  - Combination with standard chemotherapy
- **Expected Outcome**: Potent inhibition in EGFR-altered lines
- **Tool Compounds**: Erlotinib, Gefitinib, Osimertinib (all FDA-approved)
- **Timeline**: 1-2 weeks

**3. WEE1**
- **Essentiality**: Synthetic lethal with TP53 loss, selective in TP53-mutant cancers
- **Validation Approach**:
  - Test adavosertib (WEE1 inhibitor) ± DNA damaging agents
  - Stratify by TP53 status (mutant vs WT)
  - Cell cycle analysis (premature mitotic entry)
- **Expected Outcome**: Selective killing of TP53-mutant cells + synergy with chemo
- **Tool Compounds**: Adavosertib (AZD1775), PD-166285
- **Timeline**: 2-3 weeks

##### Tier 2 Targets (AURKA, CDK2, CHEK1, PLK1)

**General Strategy**:
- Pharmacological validation with 2-3 selective inhibitors per target
- Orthogonal genetic validation (CRISPRi/shRNA)
- Pathway analysis (Western blots for downstream effectors)
- Combination screens with standard-of-care agents

**Recommended Tool Compounds**:
- **AURKA**: Alisertib (MLN8237), Aurora A Inhibitor I
- **CDK2**: Roscovitine (seliciclib), Dinaciclib
- **CHEK1**: Prexasertib (LY2606368), AZD7762
- **PLK1**: Volasertib (BI 6727), BI 2536

##### Tier 3 Targets - Alternative Validation Strategies

**E2F1, RB1** (Transcription factors):
- **Challenge**: No direct small molecule inhibitors
- **Strategy**:
  - Test PROTACs if available
  - Indirect validation via upstream targets (CDK4/6 inhibitors for RB pathway)
  - Genetic validation only (CRISPRko, CRISPRi)

**MCM2-7 Complex** (Helicase, pan-essential):
- **Challenge**: Pan-cancer essential, poor therapeutic window
- **Strategy**: Deprioritize for drug development
- **Note**: Interesting for understanding replication biology, but not ideal therapeutic target

#### Validation Timeline

| Phase | Duration | Experiments | Deliverable |
|-------|----------|-------------|-------------|
| **Phase 1** | Weeks 1-3 | Tier 1 target validation (KRAS, EGFR, WEE1) | Dose-response curves, potency data |
| **Phase 2** | Weeks 4-6 | Tier 2 target validation (AURKA, CDK2, etc.) | Hit confirmation, selectivity data |
| **Phase 3** | Weeks 7-10 | Mechanism studies, pathway analysis | Western blots, cell cycle, apoptosis |
| **Phase 4** | Weeks 11-14 | Combination studies, in vivo pilot (top 2-3) | Synergy matrices, xenograft data |

#### Success Criteria for Validation

✅ **Hit Confirmed** if:
- Pharmacological inhibition phenocopies CRISPR knockout (≥50% growth inhibition at ≤1 µM)
- Dose-response curve shows IC50 consistent with essentiality score
- Effect is selective (active in screen cell line, inactive in control lines)
- Orthogonal genetic methods (siRNA, CRISPRi) reproduce phenotype

❌ **Hit Rejected/Deprioritized** if:
- Tool compounds show no effect despite strong CRISPR score (off-target CRISPR effect)
- Pan-cancer essential with no selectivity (poor therapeutic window)
- No druggable domain/strategy (TFs, scaffolds without chemical matter)
- Cannot be validated with available reagents

#### Resource Requirements

**Reagents**:
- Chemical compounds: $5-10K (tool compounds, commercial inhibitors)
- CRISPRi/shRNA validation: $3-5K (vectors, reagents)
- Cell culture & assays: $8-12K (plates, reagents, media)

**Equipment**:
- Cell culture facility (BSL2)
- Plate readers (viability assays, luminescence)
- Flow cytometer (cell cycle, apoptosis)
- Western blot equipment

**Personnel**: 1 postdoc + 1 technician for 3-4 months

---
**Validation Strategy Summary**: Focus validation efforts on Tier 1 targets (KRAS, EGFR, WEE1) with approved/late-stage drugs. These offer fastest path to clinical translation and have highest probability of success based on multi-dimensional scoring.

Report Template (Initial File)

File: CRISPR_screen_analysis_[CONTEXT].md

# CRISPR Screen Analysis Report: [CONTEXT]

**Generated**: [Date]
**Input**: [Gene list / Cancer type / Single gene]
**Context**: [Screen type, cell line, experimental details]
**Status**: In Progress

---

## Executive Summary
[Analyzing...]
<!-- Will contain: key findings, top hits, recommended priorities -->

---

## Input Validation
[Analyzing...]
<!-- Gene symbol validation, invalid genes, suggestions -->

---

## 1. Gene Essentiality Analysis
### 1.1 Strongly Essential Genes
[Analyzing...]
### 1.2 Selectively Essential Genes
[Analyzing...]
### 1.3 Weakly Essential / Non-Essential
[Analyzing...]

---

## 2. Pathway & Functional Enrichment
### 2.1 Pathway Enrichment (Reactome, KEGG)
[Analyzing...]
### 2.2 GO Enrichment (BP, MF, CC)
[Analyzing...]
### 2.3 Hallmark Gene Sets (MSigDB)
[Analyzing...]
### 2.4 Pathway-Level Interpretation
[Analyzing...]

---

## 3. Protein Interaction Network
### 3.1 Network Statistics
[Analyzing...]
### 3.2 Protein Complexes
[Analyzing...]
### 3.3 Hub Genes
[Analyzing...]
### 3.4 Synthetic Lethal Candidates
[Analyzing...]

---

## 4. Druggability Assessment
### 4.1 Target Development Level (Pharos)
[Analyzing...]
### 4.2 Approved Drugs & Clinical Candidates
[Analyzing...]
### 4.3 Chemical Probes
[Analyzing...]
### 4.4 Non-Druggable Hits (Alternative Strategies)
[Analyzing...]

---

## 5. Clinical Relevance
### 5.1 Cancer Genomic Alterations (COSMIC)
[Analyzing...]
### 5.2 Expression in Tumor vs Normal
[Analyzing...]
### 5.3 Prognostic/Predictive Biomarkers
[Analyzing...]
### 5.4 Mouse Models & Genetic Evidence
[Analyzing...]

---

## 6. Hit Prioritization & Validation
### 6.1 Multi-Dimensional Scoring
[Analyzing...]
### 6.2 Top 10 Priority Targets
[Analyzing...]
### 6.3 Validation Experiment Recommendations
[Analyzing...]
### 6.4 Validation Timeline & Resources
[Analyzing...]

---

## 7. Data Sources & Quality Control
### 7.1 Primary Data Sources
[Will be populated...]
### 7.2 Tool Call Summary
[Will be populated...]
### 7.3 Limitations & Caveats
[Will be populated...]

---

## Appendix: Full Hit List
[Complete gene list with all scores...]

When NOT to Use This Skill

• Drug target research (single gene) → Use tooluniverse-target-research skill instead
• Disease-centric query → Use tooluniverse-disease-research skill
• Chemical compound screening → Different workflow needed
• RNA-seq differential expression → Use differential expression analysis workflows
• Single gene lookup → Call DepMap tools directly

Use this skill when you have a gene list from a CRISPR screen and need comprehensive functional interpretation + target prioritization.

Example Queries That Trigger This Skill

✅ Gene List Analysis:

• “Analyze these CRISPR screen hits: EGFR, KRAS, WEE1, PLK1, AURKA, …”
• “I have 50 dropout genes from a CRISPR screen in lung cancer cells, what should I validate?”
• “Prioritize these genes for drug target development: [gene list]”

✅ Cancer Type Query:

• “What are the top essential genes in pancreatic cancer?”
• “Find druggable dependencies in triple-negative breast cancer”

✅ Single Gene Validation:

• “Is KRAS a good therapeutic target for lung cancer?”
• “Assess the druggability of WEE1 as a cancer target”

❌ Not Appropriate:

• “What is the function of EGFR?” → too broad, use target-research skill
• “Find drugs for lung cancer” → disease-centric, use drug-repurposing skill
• “Analyze this RNA-seq data” → different analytical workflow

Key Improvements from Existing Skills

Based on patterns in tooluniverse-target-research and tooluniverse-drug-research:

1. Multi-dimensional scoring system (novel for CRISPR analysis)
2. Validation experiment recommendations with timelines and reagents
3. Tier-based prioritization (Tier 1/2/3 based on actionability)
4. Tool compound suggestions for each druggable target
5. Synthetic lethal candidate identification from PPI network
6. Explicit selectivity analysis (pan-cancer vs tissue-selective)
7. Success/failure criteria for validation experiments

Quality Control Checklist

Before finalizing report:

• All input genes validated against DepMap registry
• Essentiality scores retrieved for all valid genes
• Pathway enrichment performed (minimum: GO BP, Reactome, Hallmark)
• PPI network constructed with interaction counts
• Druggability assessed for all hits (Pharos TDL + DGIdb)
• Top 10 priority targets table completed
• Validation recommendations provided for Tier 1 targets
• Evidence grades assigned (★★★, ★★☆, ★☆☆)
• All data sources cited explicitly
• “No data” explicitly stated when tools return empty results
• Executive summary synthesizes all findings