NeKo MCP Server

Model Context Protocol Integration for Biological Network Construction

This is a Model Context Protocol (MCP) server that exposes NeKo network construction capabilities to Large Language Models (LLMs). It enables LLMs to build biological networks from gene lists through natural language interactions and seamless tool chaining.

What is an MCP Server?

Model Context Protocol (MCP) is a standardized way to connect LLMs with external tools and data sources. This server:

• Exposes NeKo functionality as callable tools for LLMs
• Enables natural language network construction from gene lists
• Provides automated pathway database access through simple prompts
• Supports complex workflow chaining with downstream analysis tools

LLM Integration Patterns

1. Direct Network Construction

LLMs can build networks through natural language requests:

User Prompt: "Build a cancer network from TP53, MYC, and VEGFA using pathway databases"

LLM Tool Chain:

1. create_network(['TP53', 'MYC', 'VEGFA'], database='omnipath', max_len=2)
2. network_dimension()  # Check connectivity
3. export_network('bnet')  # Prepare for Boolean modeling

2. Database-Driven Exploration

User Prompt: "I need a comprehensive hypoxia response network. What genes should I include?"

LLM Response Pattern:

• Suggests relevant gene sets (HIF1A, VEGFA, EPAS1, etc.)
• Constructs network using pathway databases
• Analyzes connectivity and suggests improvements
• Exports in appropriate format for downstream analysis

3. Cross-Server Workflow Initiation

User Prompt: "Create a Boolean model of cell cycle regulation and simulate its dynamics"

Multi-Server Tool Chain:

NeKo Server:    create_network(cell_cycle_genes) → export_network('bnet')
MaBoSS Server:  bnet_to_bnd_and_cfg() → run_simulation()
PhysiCell:      add_physiboss_model() → multiscale_simulation()

Tool Categories Exposed

Network Construction

• create_network() - Build networks from gene lists using pathway databases
• find_paths() - Discover connections between specific genes
• Automatic database selection and parameter optimization

Network Analysis

• network_dimension() - Get topology statistics and connectivity metrics
• Built-in validation for disconnected components and optimization suggestions

Export Functions

• export_network() - Generate SIF files for analysis or BNET files for Boolean modeling
• Seamless integration with MaBoSS and PhysiCell workflows

Prompt Engineering Patterns

Pattern 1: Gene-to-Network Automation

"Build a [pathway/disease] network from these genes: [gene_list]"
→ LLM automatically: selects_database → constructs_network → validates_connectivity → exports

Pattern 2: Exploratory Network Building

"What would a comprehensive apoptosis network look like?"
→ LLM suggests: gene_sets → constructs_multiple_variants → compares_connectivity → recommends_best

Pattern 3: Workflow Pipeline Initialization

"I want to model how hypoxia affects cell fate decisions"
→ LLM chains: NeKo_network_construction → MaBoSS_boolean_simulation → PhysiCell_multiscale_modeling

Integration Benefits for LLMs

1. Database Access: LLMs gain access to comprehensive pathway databases (OmniPath, SIGNOR, Reactome)
2. Automatic Parameter Tuning: Smart suggestions for path length, database selection, and filtering
3. Error Recovery: Built-in guidance for gene symbol correction and connectivity optimization
4. Workflow Initiation: Starting point for complex Boolean and multiscale modeling pipelines
5. Biological Validation: Automatic detection of disconnected networks with improvement suggestions

Example LLM Conversation Flow

User: "I'm studying drug resistance in cancer. Help me build a relevant network."

LLM with MCP Access:

1. Suggests gene set: "I'll focus on key resistance genes: TP53, MDR1, MYC, PIK3CA, EGFR"
2. Calls: create_network(['TP53', 'MDR1', 'MYC', 'PIK3CA', 'EGFR'], database='omnipath')
3. Calls: network_dimension() to check connectivity
4. Analyzes results: "The network has 23 nodes and 45 edges, showing strong connectivity"
5. Suggests next steps: "Would you like me to export this for Boolean simulation or extend it with additional resistance pathways?"

Error Recovery and Optimization

Gene Symbol Correction

• Automatic detection of invalid gene symbols
• Suggestions for correct HGNC nomenclature
• Common alias resolution (e.g., p53 → TP53)

Connectivity Optimization

• Database selection recommendations for sparse networks
• Parameter tuning suggestions (max_len, only_signed)
• Alternative gene suggestions for disconnected components

Workflow Guidance

• Export format recommendations based on intended downstream analysis
• Integration tips for MaBoSS Boolean modeling
• PhysiCell multiscale modeling preparation

Technical Architecture

• Protocol: Model Context Protocol (MCP) standard
• Interface: JSON-RPC tool calling
• Database Integration: Real-time access to pathway databases
• Error Handling: Structured responses with actionable recovery suggestions
• Cross-Server Coordination: Seamless handoff to MaBoSS and PhysiCell servers

Getting Started

1. Install MCP client in your LLM environment
2. Connect to NeKo MCP server endpoint
3. Use natural language to request network construction
4. Chain with other MCP servers for complete modeling workflows

Learn More About NeKo: NeKo Official Documentation

This MCP server transforms NeKo from a command-line network construction tool into an LLM-accessible biological database interface, enabling natural language-driven network biology and systems modeling workflows.