Beyond Diffusion: Bayesian Learning Strategies in Single-Cell Life

A comprehensive research framework for investigating predictive learning, memory formation, and Bayesian inference at the cellular level

This repository contains the complete codebase, experimental platforms, and documentation for the paper "Beyond Diffusion: Bayesian Learning Strategies in Single-Cell Life". The project demonstrates that single-celled organisms might exhibit sophisticated learning behaviors that extend beyond simple chemical diffusion, employing Bayesian-like inference strategies for environmental prediction and adaptive responses.

• Overview
• Repository Structure
• Research Components
  • 1. MBS Simulation
  • 2. Microscope-Mounted Chemostat
  • 3. Smart Incubator
  • 4. Manuscript
• Key Findings
• Getting Started
• Research Applications
• Hardware Platforms
• Citation
• License
• Contact

This research investigates whether single-celled organisms possess the capacity for predictive learning through Bayesian-like inference mechanisms that operate at sub-neural scales. The project combines:

1. Agent-Based Simulations: Computational models comparing Markovian-Bayesian Agents (MBAs) with memory-based learning against Blind Agents (BAs) that react only to current stimuli
2. Experimental Validation: Two custom-built automated platforms for long-term, high-precision biological experiments
3. Theoretical Framework: Mathematical models of cellular learning grounded in Bayesian inference and information theory

Single-celled organisms can:

• Learn temporal associations between environmental cues and subsequent stressors
• Form cellular memories that influence future behavioral responses
• Predict environmental changes using past experience encoded in molecular states
• Optimize fitness through anticipatory rather than purely reactive strategies

MCMC-Bayesian-Learning-Strategies/
├── README.md                           # This file
├── Manuscript/
│   └── Draft_18_09_25.tex             # LaTeX manuscript source
├── MBS_simulation/
│   └── MBA vs BA sim/                 # Agent-based simulation framework
│       ├── README.md                  # Simulation documentation
│       ├── mba_vs_ba_sim/            # Core simulation package
│       ├── wrappers/                  # Experiment wrappers
│       └── Fig_pub/                   # Publication figure generation
├── Microscope_mounted_chemostat/
│   ├── README.md                      # Platform documentation (400+ lines)
│   ├── Firmware/                      # ESP32 MicroPython code
│   ├── Hardware/                      # PCB Gerbers + 3D models
│   └── Docs/                          # Specifications & methods
└── Smart_incubator/
    ├── README.md                      # Platform documentation (400+ lines)
    ├── Firmware/                      # ESP32 MicroPython code (21 files)
    ├── Hardware/                      # PCB schematics + 3D models
    └── Docs/                          # Specifications, methods & analysis

Location: MBS_simulation/MBA vs BA sim/

Purpose: Agent-based simulations comparing learning strategies in evolutionary contexts.

• Markovian-Bayesian Agents (MBAs): Organisms that maintain memory of environmental patterns and use Bayesian inference to predict future conditions
• Blind Agents (BAs): Memory-less organisms that respond only to immediate stimuli
• Moran Process: Population dynamics simulation with fitness-dependent reproduction
• Multiple Experimental Paradigms: Stress tests, genetic lock-in, continuous sweeps, topology scans

• Population Evolution: Track strategy dominance over thousands of generations
• Fitness Landscapes: Analyze preparatory response costs and benefits
• Parameter Sweeps: Systematic exploration of correlation, memory cost, environment variability
• Statistical Analysis: Permutation tests, effect size calculations, publication-ready visualizations

cd "MBS_simulation/MBA vs BA sim"
python -m wrappers.vanilla    # Basic MBA vs BA comparison
python -m wrappers.stress     # Stress test scenarios

Documentation: See MBS_simulation/MBA vs BA sim/README.md for detailed usage and CLEANED_REPOSITORY_SUMMARY.md for architecture overview.

Location: Microscope_mounted_chemostat/

Purpose: Real-time microfluidic platform for studying habituation, valence reassignment, and threshold adaptation under microscopic observation.

Experimental Control:

• 4-Pump Peristaltic System: Independent control of 4 culture chambers
• Variable Flow Rates: 0-100% PWM control (13-bit resolution)
• Programmable Protocols: Chemostat, turbidostat, morbidostat modes
• Temporal Precision: Microsecond-level timing for stimulus delivery
• Long-Term Stability: Continuous operation for days to weeks

Operational Modes:

1. Chemostat Mode: Constant dilution rate, steady-state growth
2. Turbidostat Mode: Optical density (cellpose) feedback control
3. Morbidostat Mode: Dynamic drug concentration adjustment
4. Custom Protocols: User-defined temporal sequences

Real-Time Integration:

• Microscope Compatibility: Mounts directly on inverted microscopes
• Cellpose Integration: Optional AI-based cell counting
• MQTT Communication: Remote monitoring and control
• Live Imaging: Synchronized with image acquisition systems

• Habituation Studies: Long-term adaptation to repeated chemical stressors
• Valence Reassignment: Testing whether organisms reassign meaning to previously neutral cues
• Threshold Adaptation: Investigating changes in stress response thresholds
• Chemotaxis Experiments: Studying gradient-following behavior
• Drug Resistance Evolution: Real-time observation of resistance emergence

# 1. Upload firmware to ESP32
cd Microscope_mounted_chemostat/Firmware
# Use Thonny or ampy to upload *.py files

# 2. Calibrate pumps
python local_pump_calibration.py

# 3. Run experiment
python chemostat_controller.py

Full Documentation: See Microscope_mounted_chemostat/README.md (400+ lines) and Docs/parts_list.md for complete BOM.

Location: Smart_incubator/

Purpose: Autonomous platform for testing predictive learning through controlled temporal associations between environmental cues and thermal stressors.

Thermal Control System:

• Dual Actuators: PTC heater + TEC1 Peltier cooler
• PID Regulation: Custom-tuned controller (Kp=6.0, Ki=0.02, Kd=1.5)
• Temperature Range: 15-40°C (software limited for biological safety)
• Precision: ±0.5°C at basal (23°C), ±0.8°C during heat shock (32°C)
• Rapid Transitions: 2-3 min heating, 5-8 min cooling

Multi-Modal Stimulus Delivery:

• Optical (LED): High-brightness white LED with 16-bit PWM control
• Mechanical (Vibration): Programmable vibration motor with pulsing patterns
• Combined Modes: Independent or simultaneous multi-modal delivery
• Temporal Precision: Microsecond-level timing accuracy

Experimental Protocol Framework:

Data Management:

• 10-Second Sampling: Complete system state snapshots
• SD Card Logging: FAT32 with SHA-256 checksums
• Structured Hierarchy: Experiment directories with metadata
• Data Integrity: Cryptographic verification and manifest tracking

Safety Features:

• Temperature Failsafe: Stuck sensor detection and overheat protection
• Automated Recovery: Sensor reinitialization and error resilience
• Progressive Protection: Graduated intervention based on severity
• Memory Management: Periodic garbage collection and heap monitoring

• Predictive Learning: Testing temporal association between cues and stress
• Memory Formation: Examining cellular response persistence
• Phenotypic Plasticity: Multi-timescale environmental adaptation
• Pre-Neural Learning: Investigating sub-neural inference mechanisms
• Complex Protocols: Custom landscape execution with multi-parameter control

# 1. Format SD card as DOS_FAT_32 with MBR
diskutil eraseDisk MS-DOS INCUBATOR MBR /dev/diskX

# 2. Upload firmware to ESP32
cd Smart_incubator/Firmware
# Use Thonny or ampy to upload *.py files

# 3. Power on - system auto-initializes and starts cycles
# Monitor via OLED display or serial console

Full Documentation: See Smart_incubator/README.md (400+ lines) and Docs/parts_list.md for complete BOM.

For Simulations:

# Python 3.8+ with scientific computing stack
pip install numpy scipy matplotlib pandas seaborn

For Hardware Platforms:

• ESP32 development boards
• MicroPython firmware (v1.24.1+)
• Electronic components (see respective parts lists)
• Basic electronics assembly skills

Clone Repository:

git clone https://github.com/your-repo/MCMC-Bayesian-Learning-Strategies.git
cd MCMC-Bayesian-Learning-Strategies

Run Simulations:

cd "MBS_simulation/MBA vs BA sim"
python -m wrappers.vanilla

Build Hardware:

1. Review platform-specific README files for detailed instructions
2. Order components from parts_list.md BOMs
3. Follow assembly guides in respective Docs/ directories
4. Upload firmware and configure for your experiments

Both platforms share:

• ESP32-based control with MicroPython
• PWM actuation for precise analog control
• SD card data logging with integrity verification
• Real-time monitoring (MQTT/OLED display)
• Safety systems (sensor recovery, failsafes)
• Modular design for easy modification
• Open-source firmware and hardware designs

If you use this work in your research, please cite:

@article{your_paper_2025,
  title={Beyond Diffusion: Bayesian Learning Strategies in Single-Cell Life},
  author={[Authors]},
  journal={[Journal]},
  year={2025},
  doi={[DOI]}
}

[Specify license - e.g., MIT, GPL, CC-BY, etc.]

This repository contains:

• Software: Simulation code and firmware (specify license)
• Hardware: PCB designs and 3D models (specify license, e.g., CERN-OHL)
• Documentation: Guides and specifications (specify license, e.g., CC-BY-4.0)

Institution: [IBE, CSIC Barcelona, MCG Lab ]

We welcome contributions! Please:

1. Fork the repository
2. Create a feature branch
3. Submit a pull request with clear description

For hardware modifications, include:

• Updated BOMs
• Revised schematics/PCB files
• Testing notes and validation data

• Issues: Report bugs via GitHub Issues
• Discussions: Use GitHub Discussions for questions
• Documentation: Refer to platform-specific README files
• Hardware: See respective parts_list.md for component sourcing

Current Version: 1.0 (September 2025)

Status:

• ✅ Simulations: Complete and validated
• ✅ Microscope-Mounted Chemostat: Built and operational
• ✅ Smart Incubator: Built and operational
• 🔄 Manuscript: Under review

Roadmap:

• Extended simulation parameter sweeps
• Additional experimental validation datasets
• Protocol library expansion
• Community hardware builds and feedback

Repository Structure Last Updated: September 30, 2025
Maintainers: [Maor Knafo]
For questions or collaboration inquiries: [Contact info]