EnzBuilder: Simultaneously Improving Metabolic Model Reconstruction and Enzyme Function Annotation with Machine Learning Methods

EnzBuilder integrates predicted protein annotations directly into the metabolic reconstruction and refinement processes.

---

Table of Contents

1. About
2. Features
3. Installation
4. Usage
5. CLEAN Prediction
6. Contributing

---

About

EnzBuilder is a tool designed to bridge the gap between protein annotation and metabolic model reconstruction by leveraging machine learning methods.

• Incorporates predicted enzyme functions into draft models.
• Enables iterative model refinement using AI predictions.
• Enhances accuracy and coverage of metabolic networks.

---

Features

• Seamless integration of machine learning-based protein annotations.
• Automated model gap-filling based on functional predictions.
• Compatibility with common COBRA and SBML workflows.
• Iterative refinement of metabolic models to improve network coverage.

---

Installation

To install and set up EnzBuilder, follow these steps:

# Clone the repository
git clone https://github.com/yourusername/EnzBuilder.git

# Navigate to the project directory
cd EnzBuilder

# Install dependencies
conda create --name enzbuilder --file enzbuilder.yml

# Activate the conda environment
conda activate enzbuilder

Data Download

1. Download the required data file from Google Drive.
2. Extract the data by running:

cd data
tar -xzvf data.tar.gz

---

Usage

Basic metabolic model reconstruction

Run the following command to process your input file:
test input file can be found in the previous data.tar.gz

python funcarve_main.py \
  --input_file ../data/test/CP000148.1_t4_maxsep_df.pkl \
  --gram negative \
  --block_flage 0 \
  --flux_flage 0 \
  --name iLJ478_allec \
  --media default \
  --reward 0.1 \
  --iter 3 \
  --cpu 8 \
  --threshold 5 \
  --upper 15 \
  --lower 5 \
  --maxweight 100 \
  --minweight 0.0

Parameter explanation

• --input_file: Path to the input file. Default is 'none'.
• --file_type: Input file type. Options are cleandf=1 or fasta=2. Default is 1.
• --reward: Reward for new reactions to update predscore. Default is 0.1.
• --iter: The number of iterations to run the funcarve algorithm. Default is 3.
• --block_flage: Decrease block reactions (1) or ignore block reactions (0). Default is 1.
• --flux_flage: Reward changed by flux of the reactions (1) or ignore reactions fluxes (0). Default is 0.
• --media: List of metabolites composing the media condition. Not required. Default is 'rich'.
• --tasks: List of metabolic tasks. Not required. Default is an empty list.
• --org: KEGG organism code. Not required. Default is 'default'.
• --min_frac: Minimum objective fraction required during gapfilling. Default is 0.01.
• --max_frac: Maximum objective fraction allowed during gapfilling. Default is 0.5.
• --threshold: The cutoff value for the EC prediction score. Default is 5.
• --upper: Upper threshold for predscore. Default is 15.
• --lower: Lower threshold for predscore. Default is 5.
• --maxweight: Maximum weight for reactions. Default is 100.
• --minweight: Minimum weight for reactions. Default is 0.0.
• --gram: Type of Gram classification (positive or negative). Default is 'none'.
• --out: Name of output GENRE file. Default is 'defaultpath'.
• --name: ID of output GENRE. Default is 'default'.
• --cpu: Number of processors to use. Default is 1.
• --gapfill: Gapfill your model? Default is 'yes'.
• --exchange: Open exchange (1) or shut down exchange (0). Default is 1.
• --startindex: The start index of the iteration. Default is 1.

---

CLEAN Prediction

EnzBuilder relies on CLEAN for enzyme function predictions. Follow these steps to set up CLEAN:

1. Install CLEAN by following the instructions at CLEAN GitHub.
2. Replace CLEAN/app/src/CLEAN/infer.py with the version in src/infer.py from EnzBuilder.
3. When running CLEAN, an additional .pkl file will be generated. Use this file as input for EnzBuilder.

---

Contributing

We welcome contributions! If you'd like to contribute, please fork the repository, create a branch, and submit a pull request.

For bug reports or feature requests, please open an issue on the GitHub repository.