m6a-detection-project

Table of Contents

• Overview
  
• Repository Structure
  
• Quick Start Guide
  
  • Ubuntu setup
    
  • Cloning the repository
    
  • Installing dependencies
    
• Usage
  
• Interpretation of Outputs
  
• Citing
  
• License
  

Overview

This project aims to address challenges that exist in developing a robust machine-learning classifier by developing a CNN model for identifying m6A modifications in RNA-Seq data, specifically focusing on cell lines found in the SG-NEx Project (2021). By enhancing our ability to detect m6A modifications, we hope to contribute to a deeper understanding of identification of m6A and its potential as a target for therapeutic intervention.

Repository Structure

This section provides an overview of the main folders and files in this repository. bash m6a-detection-project/ ├── data/ # Folder containing RNA-Seq data files │ └── test_data.json # Test data used for running model tests ├── output/ # Folder where model prediction files are saved │ └── model_output_datetime.csv # Example output file with prediction results │ ├── test.py # Script for evaluating the model on test data ├── utils.py # Script containing all relevant functions used in the testing script ├── train.ipynb # Notebook for training and tuning the model ├── final_task2.ipynb # Notebook for testing model to predict m6A sites in all SG-NEx direct RNA-Seq data sets ├── cnn_selected.h5 # Final trained model file │ ├── CITATION.cff # Citation for this repository ├── LICENSE # License for this repository ├── requirements.txt # List of required packages and dependencies ├── README.md # Project overview and setup instructions (this file) └── .gitignore # Specifies files and folders to ignore in version control

Quick Start Guide

Ubuntu setup

1. Start an Ubuntu instance from ResearchGateway. A recommended and sufficient Ubuntu instance is: t3.medium.
2. Access your Ubuntu instance.

Cloning the repository

1. Clone the repository. To clone our repository using HTTPS, run bash git clone https://github.com/louisetxz/m6a-detection-project.git
2. Change directory into our project. cd m6a-detection-project

Installing dependencies

To update package list, run bash sudo apt-get update To install package manager PIP, run bash sudo apt install python3-pip To install dependencies, run bash pip install -r requirements.txt Note: These dependencies are required only for the testing file, test.py.

Usage

There are two ways you can generate predictions using our model:

• To generate predictions on the preloaded test data available in the repository, run: bash python3 test.py

• To generate predictions with your own data or your own command-line arguments:

  1. Place your data under data/. The data should follow the format of the direct RNA-Seq data as in the SG-NEx project.

2. Run:
```bash
python3 test.py --model_path /path/to/model --data_path /path/to/data --n 5 --output_filename your-filename.csv
```

The testing script contains the following command-line arguments:
* --model_path (str): Path to the trained model file. Default is cnn_selected.h5.

* --data_path (str): Path to the test dataset file which must be in json format. Default is /data/test_data.json.

* --n (int): Number of rows of the predictions to print to the console. Default is 10.

* --output_filename (str): File name for the output predictions. This will save the results to the specified file in the /output folder. Default is model_output_`datetime`.csv, where `datetime` is captured from the server where you are running the code. In an Ubuntu instance, it will follow the UTC timezone.

Interpretation of outputs

The predictions will be saved as a csv file (Default name: model_output_datetime.csv) which can be found under the /output folder. It contains the results of modification at each individual position for each transcript. To see the predictions, follow these steps:

1. Navigate to the output directory to locate the file with the prediction results: bash cd output
2. List all files in the output directory to identify the latest predictions file: bash ls
3. Display the predictions in the terminal by specifying the appropriate filename (e.g., modeloutputdatetime.csv): bash cat model_output_datetime.csv # Replace with the correct output file name

The output file will the following three columns:

• transcript_id: The transcript id of the predicted position
• transcript_position: The transcript position of the predicted position
• score: The probability that a given site is modified

Citing

If you use this model in your research, please cite this repository: @misc{Louise_Tan_Xuan_Zhi_and_Khine_Ezali_and_Lim_Shih_Ler_Sean_and_Sitoh_Ying_Ting_Rachel_m6a-detection-project_2024, author = {Louise Tan Xuan Zhi and Khine Ezali and Lim Shih Ler Sean and Sitoh Ying Ting Rachel}, license = {MIT}, month = nov, title = {{m6a-detection-project}}, url = {https://github.com/louisetxz/m6a-detection-project}, version = {1}, year = {2024} } Or cite Louise Tan Xuan Zhi, Khine Ezali, Lim Shih Ler Sean, & Sitoh Ying Ting Rachel. (2024). m6a-detection-project (Version 1) [CNN Model]. https://github.com/louisetxz/m6a-detection-project

License

m6Anet is licensed under the terms of the MIT license.