Neural-Bio: Hybrid Modeling Framework for Bioreactor Optimization

A hybrid modeling framework combining neural networks with physics-based constraints for bioreactor process optimization and control.

• Neural network for reaction rate prediction
• First principle model for bioreactor dynamics
• Physics-informed constraints integration
• Adaptive weighted loss function

Overview

🧬 Neural-Bio is a PyTorch-based framework that implements a hybrid modeling approach for bioprocess optimization. It combines deep learning with mechanistic models to predict and optimize bioreactor processes, incorporating both data-driven insights and physical constraints.

Installation

# Clone the repository
git clone https://github.com/deepbiolab/neural-bio.git
cd neural-bio

# Create a virtual environment (optional but recommended)
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Usage

Training

python main.py --mode train

• Place training data files in the dataset/train folder
• Ensure data follows the required format (see Data Format section)
• Adjust configuration settings in src/config/default_config.py as needed
• Training and validation performance imgs are saved in the logs/ directory

Testing

python main.py --mode test

• Place test data files in the dataset/test folder
• Ensure data follows the required format (see Data Format section)
• Verify model checkpoint exists in checkpoints/ directory
• Test performance is saved in the logs/ directory

Prediction

python main.py --mode predict

• Place input data files in the dataset/predict folder
• Ensure data follows the required format (see Data Format section)
• Verify model checkpoint exists in checkpoints/ directory
• Prediction results are saved in the results/ directory

Project Structure

neural-bio/
├── src/
│   ├── config/          # Configuration settings
│   ├── data/            # Data processing and dataset classes
│   ├── models/          # Neural network and hybrid model implementations
│   ├── trainer          # Training utilities
│   ├── inference        # inference utilities
│   └── utils/           # Visualization and evaluation tools
├── dataset/             # Data directory (not included)
├── checkpoints/         # Model checkpoints
├── results/             # Prediction results
├── logs/                # Training logs
└── main.py              # Main script for training, prediction, and testing

Configuration

The model behavior can be customized through the configuration file src/config/default_config.py. Key parameters include:

• Dataset parameters (time steps, variables, etc.)
• Model architecture (layers, dimensions)
• Training parameters (learning rate, batch size)
• Loss function weights
• Evaluation metrics

Data Format

The framework expects two main data files in CSV format:

OWU (Observation Wise Unit) Data

Time-series measurements of process variables for each experimental run.

run,time,VCD,Glc,Lac,Titer,Glc_feed

• run: Experiment run index
• time: Time point (days)
• State Variables:
  • VCD: Viable Cell Density (10⁶ cells/mL)
  • Glc: Glucose concentration (g/L)
  • Lac: Lactate concentration (g/L)
  • Titer: Product titer (g/L)
• Feeding Variables:
  • Glc_feed: Glucose feeding rate (g/L/day)

DOE (Design of Experiments) Data

Experimental design parameters and initial conditions for each run.

run,feed_start,feed_end,Glc_feed_rate,Glc_0,VCD_0

• run: Experiment run index
• Feeding Schedule:
  • feed_start: Day to start feeding
  • feed_end: Day to stop feeding
  • Glc_feed_rate: Glucose feed concentration
• Initial Conditions:
  • Glc_0: Initial glucose concentration
  • VCD_0: Initial viable cell density

Data Organization

dataset/
├── interpolation/
│   ├── train/
│   │   ├── owu.csv
│   │   └── owu_doe.csv
│   ├── test/
│   │   ├── owu.csv
│   │   └── owu_doe.csv
│   └── predict/
│       ├── owu.csv
│       └── owu_doe.csv

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Citation

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

@software{neural_bio2025,
  author = {Tim-Lin},
  title = {Neural-Bio: Hybrid Modeling Framework for Bioreactor Optimization},
  year = {2025},
  publisher = {GitHub},
  url = {https://github.com/deepbiolab/neural-bio}
}

Contact

For questions and feedback, please open an issue on GitHub.