MCP Capability Assessment: Supported Query Types and Limitations

[turbomam]

MCP Capability Assessment: Supported Query Types and Limitations

This issue documents the types of queries the fitness MCP should be able to handle effectively, organized by complexity and tool requirements.

✅ Single Tool Queries (Should Handle Easily)

Basic Information Lookup

• "What is gene Atu0001?"
• "Tell me about condition set10IT004"
• "What does a fitness score of -0.8 mean?"
• "List all available growth conditions"
• "Show me all functional modules"

Simple Searches

• "Find genes related to ribosome"
• "Search for modules containing 'transport'"
• "Find genes with 'RNA' in their description"
• "What conditions contain 'pH' in their name?"

Direct Analysis

• "Which genes are essential (fitness threshold 0.5)?"
• "Find genes that inhibit growth (threshold -0.5)"
• "What modules contain gene Atu0001?"
• "Which genes are in module 5?"

🔗 Multi-Tool Chain Queries (Should Handle with Tool Chaining)

Gene-Focused Analysis Workflows

• "Find all ribosomal genes, then show their fitness effects and functional modules"
  • search_genes("ribosom") → get_fitness_effects_for_gene() → get_gene_modules()
• "Analyze gene Atu0001's fitness across conditions and interpret the biological meaning"
  • analyze_gene_fitness("Atu0001") → interpret_fitness_score() for each condition
• "Find essential genes in pH stress conditions, then get their module information"
  • find_essential_genes(condition_filter="pH") → get_gene_modules() for each

Condition-Focused Analysis Workflows

• "What genes are affected by condition X, and what modules do they belong to?"
  • get_genes_with_fitness_effects("condition_X") → get_gene_modules() for each gene
• "Find all metal stress conditions, then identify essential genes in each"
  • get_growth_conditions("metal") → find_essential_genes(condition_filter=condition) for each
• "Compare fitness effects between different carbon source conditions"
  • get_growth_conditions("carbon") → get_genes_with_fitness_effects() for each condition

Network Exploration Workflows

• "Starting from gene X in condition Y, map the fitness interaction network"
  • get_fitness_effects_for_gene("X") → expand_fitness_network("X", "Y") → explore related genes/conditions
• "Find genes similar to rpoA based on fitness patterns across conditions"
  • get_fitness_effects_for_gene("rpoA") → identify conditions → get_genes_with_fitness_effects() for each condition
• "Identify functional modules that are important under stress conditions"
  • get_growth_conditions("stress") → find_essential_genes() for each → get_gene_modules() → aggregate by module

Comparative Analysis Workflows

• "Compare the fitness profiles of genes in the flagellar biosynthesis module"
  • search_modules("flagellar") → get_module_genes() → analyze_gene_fitness() for each gene
• "Find conditions where essential genes and growth inhibitors overlap"
  • find_essential_genes() → find_growth_inhibitor_genes() → cross-reference conditions

❌ Queries MCP Cannot Handle (Limitations)

External Data Requirements

• "Compare this fitness data with E. coli fitness data"
  • Limitation: No cross-organism data available
• "What's the latest research on gene Atu0001?"
  • Limitation: No literature database integration
• "Show me the protein structure of this gene"
  • Limitation: No structural data available

Advanced Statistical Analysis

• "Perform principal component analysis on all fitness data"
  • Limitation: No advanced statistical analysis tools
• "Calculate correlation matrices between all genes"
  • Limitation: Complex statistical computations not implemented
• "Run GSEA analysis on these gene sets"
  • Limitation: No gene set enrichment analysis capability

Data Modification/Experimental Design

• "Add new experimental data for condition X"
  • Limitation: Read-only data access
• "Design an experiment to test gene X under stress"
  • Limitation: No experimental design capability
• "Update the fitness threshold to 3.0"
  • Limitation: Configuration changes require code modification

Real-time/Dynamic Analysis

• "Monitor fitness changes over time"
  • Limitation: Static dataset, no time-series data
• "Predict fitness under new conditions"
  • Limitation: No predictive modeling capability
• "Simulate gene knockouts"
  • Limitation: No simulation or modeling tools

Complex Biological Interpretation

• "Explain the molecular mechanism behind this fitness effect"
  • Limitation: No mechanistic knowledge base
• "Predict protein-protein interactions from fitness data"
  • Limitation: No interaction prediction capability
• "Determine optimal growth conditions for maximum yield"
  • Limitation: No optimization algorithms

🎯 Testing Strategy

Automated Testing

• Create test cases for each single-tool query type
• Develop integration tests for common multi-tool workflows
• Add negative tests for unsupported query types

User Experience Testing

• Test query complexity escalation (simple → complex)
• Validate tool chaining suggestions work correctly
• Ensure clear error messages for unsupported queries

Performance Testing

• Benchmark single vs multi-tool query response times
• Test large-scale network expansion queries
• Validate memory usage under complex workflows

📋 Implementation Notes

• The suggestions field in tool responses should guide users toward appropriate tool chains
• Error messages should clearly distinguish between "data not found" vs "query type not supported"
• Consider adding workflow templates for common multi-tool analysis patterns
• Document performance characteristics for different query complexities