PhysiCell MCP Server

Model Context Protocol Integration for Multiscale Biological Simulation

This is a Model Context Protocol (MCP) server that exposes PhysiCell multicellular simulation capabilities to Large Language Models (LLMs). It enables LLMs to construct sophisticated physics-based tissue simulations with integrated Boolean gene networks through natural language interactions.

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 PhysiCell functionality as callable tools for LLMs
• Enables natural language simulation construction from biological scenarios
• Provides PhysiBoSS integration for multiscale gene-to-tissue modeling
• Supports complex workflow orchestration across multiple biological scales

LLM Integration Patterns

1. Scenario-Driven Simulation Building

LLMs can construct complete simulations from biological descriptions:

User Prompt: "Create a simulation of breast cancer cells in a hypoxic 3D environment with immune cell infiltration"

LLM Tool Chain:

1. analyze_biological_scenario("Breast cancer in hypoxic 3D tissue with immune infiltration")
2. create_simulation_domain(3000, 3000, 500, max_time=7200)
3. add_single_substrate("oxygen", 100000, 0.01, 38)
4. add_single_cell_type("cancer_cell", "Ki67_basic")
5. add_single_cell_type("immune_cell", "live_cell")
6. add_single_cell_rule("cancer_cell", "oxygen", "decreases", "apoptosis rate")
7. export_xml_configuration("tumor_simulation.xml")

2. Multiscale Integration Workflows

User Prompt: "Integrate the p53 Boolean network into cancer cell behavior and simulate tumor growth under drug treatment"

Cross-Server Tool Chain:

NeKo:       create_network(['TP53', 'MDM2', 'ATM']) → export_network('bnet')
MaBoSS:     bnet_to_bnd_and_cfg() → test Boolean dynamics
PhysiCell:  add_physiboss_model() → link genes to behaviors → simulate

3. Interactive Simulation Design

User Prompt: "What cell types and environmental factors should I include for studying drug resistance?"

LLM Response Pattern:

• Analyzes biological scenario
• Suggests appropriate cell types and substrates
• Recommends signal-behavior rules
• Guides PhysiBoSS integration for gene-level control
• Provides complete simulation configuration

Tool Categories Exposed

Simulation Framework

• create_simulation_domain() - Define 3D spatial and temporal boundaries
• add_single_substrate() - Add chemical environments (oxygen, drugs, nutrients)
• Session management with progress tracking across complex workflows

Cell Population Definition

• add_single_cell_type() - Define cancer, immune, stromal cell populations
• configure_cell_parameters() - Set size, motility, death rates
• set_substrate_interaction() - Define consumption and secretion

Behavioral Programming

• add_single_cell_rule() - Create environmental sensing and response
• list_all_available_signals() and list_all_available_behaviors() - Discovery tools
• Context-aware signal/behavior expansion based on simulation components

PhysiBoSS Multiscale Integration

• add_physiboss_model() - Integrate Boolean networks into cell behavior
• add_physiboss_input_link() - Connect environment to gene regulation
• add_physiboss_output_link() - Connect gene states to cell phenotypes
• apply_physiboss_mutation() - Simulate genetic perturbations

Prompt Engineering Patterns

Pattern 1: Complete Simulation from Description

"Simulate [disease/scenario] with [environmental conditions] and [cell types]"
→ LLM automatically: scenario_analysis → domain_setup → cell_definition → rule_programming → export

Pattern 2: Multiscale Model Construction

"Connect this Boolean network to cell behavior in a tissue simulation"
→ LLM chains: physiboss_integration → input_output_linking → parameter_tuning → validation

Pattern 3: Iterative Simulation Refinement

"Add drug treatment effects to my existing simulation"
→ LLM extends: add_drug_substrate → modify_behavioral_rules → update_cell_interactions

Advanced PhysiBoSS Integration

Multiscale Architecture through MCP

The server enables LLMs to seamlessly connect molecular and cellular scales:

Gene Regulation (Boolean) ↔ Cell Behavior (PhysiCell) ↔ Tissue Dynamics (3D Physics)
        ↓                         ↓                        ↓
   Input: Environment        Output: Phenotype      Emergent: Population
   (oxygen, drugs)          (death, proliferation)   (growth, invasion)

LLM-Orchestrated Multiscale Workflows

User Prompt: "Model how TP53 mutations affect tumor response to chemotherapy"

LLM Multiscale Tool Chain:

1. Network Level: Construct TP53 regulatory network (NeKo)
2. Boolean Level: Simulate pathway dynamics (MaBoSS)
3. Cellular Level: Link TP53 states to apoptosis/survival (PhysiCell PhysiBoSS)
4. Tissue Level: Simulate drug treatment effects on tumor population
5. Analysis: Compare wild-type vs mutant tumor responses

Session Management for Complex Workflows

Multi-Session Orchestration

• create_session() - Isolated simulation environments
• switch_session() - Compare different scenarios
• get_workflow_status() - Track progress across complex builds

Workflow State Tracking

LLMs can monitor and guide users through simulation construction:

• Domain setup → Substrates → Cell types → Rules → PhysiBoSS → Export
• Progress percentages and next-step recommendations
• Error recovery with specific correction guidance

Integration Benefits for LLMs

1. Biological Scenario Translation: Convert narrative descriptions into quantitative simulations
2. Multiscale Coordination: Orchestrate gene→cell→tissue modeling workflows
3. Parameter Discovery: Access to extensive signal/behavior libraries with context awareness
4. Error Prevention: Built-in validation and progress tracking
5. Cross-Server Integration: Seamless coordination with NeKo and MaBoSS servers

Example LLM Conversation Flow

User: "I want to study how hypoxia drives cancer cell invasion with p53 mutations."

LLM with MCP Access:

1. Scenario Analysis: "I'll create a hypoxic tumor simulation with p53-controlled invasion"
2. Domain Setup: Creates 3D environment with oxygen gradients
3. Cell Definition: Adds cancer cells with motility and invasion capabilities
4. PhysiBoSS Integration:
   • Links oxygen levels to HIF1A activation
   • Connects p53 mutations to survival/invasion decisions
   • Programs hypoxia-induced migration behaviors
5. Simulation Export: Generates complete XML configuration
6. Next Steps: "The simulation is ready. Would you like me to add immune cells or test different p53 mutation scenarios?"

Technical Architecture

• Protocol: Model Context Protocol (MCP) standard
• Interface: JSON-RPC tool calling with complex object handling
• State Management: Session-based simulation building with persistence
• PhysiBoSS Integration: Direct Boolean network coupling to cellular physics
• Cross-Server Coordination: Seamless file handoff from NeKo/MaBoSS workflows

Getting Started

1. Install MCP client in your LLM environment
2. Connect to PhysiCell MCP server endpoint
3. Use natural language to describe biological scenarios
4. Chain with NeKo/MaBoSS for complete gene→tissue modeling

Learn More About PhysiCell: PhysiCell Official Documentation
Learn More About PhysiBoSS: PhysiBoSS Publication

This MCP server transforms PhysiCell from a complex simulation framework into an LLM-accessible multiscale modeling platform, enabling natural language-driven construction of sophisticated gene-to-tissue simulations.