SolvIT - Protein Solubility Prediction Deep Neural Network

SolvIT is a machine learning model designed to predict protein solubility in Escherichia coli and aid in enzyme design by prioritizing high-solubility candidates from large design sets. This approach leverages a small graph neural networks (GNNs) to achieve state-of-the-art performance that rivals much larger models.

Key Features

• Deep Learning Integration: Uses SolvIT, a GNN-based solubility classifier trained on E. coli expression data.
• Comprehensive Pipeline: Automates feature extraction, solubility prediction, and result formatting.
• Ease of Use: Minimal setup requirements for running pre-defined protein designs.

---

Prerequisites

Before running the SolvIT pipeline, ensure the following tools are installed:

• Apptainer/Singularity: Used for managing the containerized environments. Installation Instructions

Additional Requirements

• Python environment specified in the environment.yaml file provided in this repository.

---

Installation

1. Clone the repository:

   git clone https://github.com/Enzymit/SolvIT.git
   cd SolvIT

2. Create and activate the Python environment:

   conda env create -f environment.yaml
   conda activate solvit_snakemake

3. Download the necessary Singularity container:

   ./singularity/download_sif.sh

---

Usage

Preliminary Configuration

Modify the config.yaml file as needed. Key parameters include:

• OUTDIR: Output directory for results.
• INPUTDIR: Directory containing input .pdb files.
• SINGULARITY_PATH: Path to the directory containing the downloaded .sif file. (usually singularity)
• OUTFILENAME: Name of the final output file.

Running the Pipeline

1. Execute the Snakemake workflow:

   snakemake --cores <number_of_cores> --use-singularity

   Replace <number_of_cores> with the number of CPU cores to use.

Output

The final results will be saved in the output directory specified in config.yaml under the name provided in OUTFILENAME.

---

Pipeline Overview

The pipeline consists of the following steps:

1. Feature Extraction: Extracts features from input .pdb files using Rosetta and saves them in a compressed format.
2. SolvIT Prediction: Runs the solubility prediction model on the extracted features.
3. Result Formatting: Processes raw predictions into a user-friendly .csv file.

Example Configuration (config.yaml)

OUTDIR: "output"
INPUTDIR: "example"
SINGULARITY_PATH: "singularity"
OUTFILENAME: "solvit_out.csv"

---

Citation

If you use SolvIT in your research, please cite:

Zimmerman et al., Context-dependent design of induced-fit enzymes using deep learning generates well-expressed, thermally stable, and active enzymes. PNAS, 2024. DOI:10.1073/pnas.2313809121

---

License

This project is licensed under the GNU General Public License v3.0 (GPL-3.0).

---

For any questions or issues, create a new issue in the repository.