Cellular Bayes Filtering for In Vivo Migration

A comprehensive simulation framework for studying cell migration in tissue environments using Bayesian filtering and biochemical navigation strategies. This repository implements both MATLAB-based cellular simulations and traditional Python-based robotics algorithms.

Key Features

• Multiple Navigation Strategies: Biochemical gradient sensing, Bayesian filtering
• Tissue Environment Simulation: Realistic tissue environments
• LEGI Models: Local Excitation, Global Inhibition models
• Robotic Implementation: Python equivalent using advanced filtering techniques (EKF)

Directory Structure

.
├── src/                    # Core MATLAB navigation simulation code
├── tissue_sim/            # Tissue environment generation and simulation
│   ├── koff_variants/     # Different binding kinetics environments
│   └── point_source/      # Point source gradient environments
├── LEGI/                  # LEGI model implementations
├── robotic/               # Python robotic implementation
│   ├── core/             # Base environment and configuration
│   ├── filters/          # EKF, UKF filter implementations
│   ├── histogram_filter/ # Bayesian histogram filter
│   └── analysis/         # Comparison and analysis tools
├── run.m                  # Main MATLAB example script
└── requirements.txt       # Python dependencies

Requirements

MATLAB Requirements

• MATLAB R2018b or later
• Statistics and Machine Learning Toolbox (for statistical functions)
• Signal Processing Toolbox (recommended)

Python Requirements

• Python 3.8+
• Dependencies listed in requirements.txt:
  • numpy, scipy, matplotlib
  • filterpy (for Kalman filtering)
  • Other scientific computing libraries

Installation

MATLAB Setup

1. Clone the repository:

   git clone <repository-url>
   cd bayesian-cell

2. Add the project directory and subdirectories to your MATLAB path:

   addpath(genpath('/path/to/bayesian-cell'))

Python Setup

1. Install Python dependencies:

   pip install -r requirements.txt

2. Run Python simulations:

   cd robotic
   python main.py

Key Functions

racing_cells Function

[schemerate,unifrate,statsummary] = racing_cells(fname, param, varargin)

Primary function for simulating cell migration in tissue environments.

Parameters:

• fname: Environment file name (e.g., "tissue_point_noflow")
• param: Simulation parameters structure
• Optional Parameters:
  • 'task': "localization" or "retention" (default: "localization")
  • 'envmodel': "tissue" or "grad" (default: "tissue")
  • 'receptor': "feedback", "uniform", or "w1dist" (default: "feedback")
  • 'decoder_method': "optimal_noise", "perfect", or "randomwalk"
  • 'source': "edge" or "point" (default: "edge")
  • 'save_data': 0 (no save), 1 (basic), 2 (all data) (default: 2)
  • 'makeplot': Enable/disable plotting (default: true)

Returns:

• schemerate: Success rate using specified navigation strategy
• unifrate: Success rate using uniform random strategy
• statsummary: Detailed statistics summary

Usage Examples

Basic MATLAB Simulation

Run the default example:

run

Custom Simulations

Bayesian vs Gradient-tracking Comparison

% Bayesian filtering approach
bayes_rate = racing_cells("tissue_env", scheme_param, ...
    "receptor", "feedback");

% Pure biochemical approach  
biochem_rate = racing_cells("tissue_env", scheme_param, ...
    "receptor", "uniform");

Environment Variants

% Different binding kinetics
racing_cells("tissue_env_koff=1e-3", scheme_param);
racing_cells("tissue_env_koff=1e-1", scheme_param);

% Point source vs edge source
racing_cells("tissue_point", scheme_param, "source", "point");

Python Robotic Implementation

cd robotic
python main.py

Tissue Environments

Available Environment Files

• tissue_env.mat: Standard tissue environment with gradient
• tissue_point.mat: Point source gradient

Environment Parameters

• Gradient strength and direction
• Binding/unbinding rates (kon, koff)
• Tissue geometry and boundary conditions
• Flow fields (where applicable)

LEGI Models

The LEGI/ directory contains Local Excitation, Global Inhibition models (code from Shi et al (2013)):

• LEGI_BEN_POL.m: LEGI model with BEN polarization
• LEGI_BEN_POL_LSM.m: LEGI with Lateral Segregation Model
• RDS_grdt_sde.m: Reaction-diffusion system with gradient and stochastic differential equations
• make_init.m: Initialize LEGI model parameters
• plot_BEN_POL.m: Visualization tools for LEGI results