BioModelComparison

This project compares the effectiveness of deep learning architectures and training protocols on various biological datasets.

Requirements

This project requires Python 3.9 or later. You can download Python from the official website. We recommend using conda to manage your Python environment. The expected size of the project is around 4.1 GB for the datasets and 5.6 GB total.

Setting up the environment

Download this project to your machine, or clone it using git.

Change (cd) into the project directory and create a new virtual environment. Using conda on macOS/Linux, you can create a new environment with the required packages by running the following commands in your terminal:

python -m venv .env
source .env/bin/activate
pip install -r requirements.txt

Downloading data

To access the datasets used in this work, you must set up a Kaggle account and download a Kaggle API key. You can do this by following the instructions here. Briefly, you must download the API key from your Kaggle account settings and place it in the ~/.kaggle/ directory.

Next you will need to install a rar extraction tool. On Linux, you can do this by running the following command in your terminal:

Linux (e.g., Ubuntu/Debian)

sudo apt-get install unrar

For macOS, you can download rar or any other unrar-ing tools. Here is an example using the Homebrew package manager:

brew install rar

Once you have your API key and rar, you can download the datasets by running the following commands in your terminal:

python datasets/download_data.py

This will take a few minutes to download the datasets and extract them into the data/ directory.

Tutorials

In this repository, we include a few tutorials that demonstrate how to train deep learning models on biological datasets. These tutorials can be found in the tutorials/ directory.