Revisiting K-mer Profile for Effective and Scalable Genome Representation Learning

Overview

This project explores effective and scalable genome representation learning approaches relying on the k-mer features for the metagenomics binning task.

Installation

1. Clone this repository:

git clone https://github.com/abdcelikkanat/revisitingkmers.git
cd revisitingkmers

2. Install dependencies: Make sure you have Python 3.8 installed. You can install the required Python packages using pip:

pip install -r requirements.txt

3. Install gdown (if you don't already have it) for downloading the datasets:

pip install gdown

Datasets

To download and prepare the training dataset, run the following commands:

gdown 1p59ch_MO-9DXh3LUIvorllPJGLEAwsUp
unzip dnabert-s_train.zip

To download the evaluation datasets, use the following commands:

gdown 1I44T2alXrtXPZrhkuca6QP3tFHxDW98c
unzip dnabert-s_eval.zip

Usage

To view the detailed usage instructions for each model, you can use the --help flag:

Poisson Model

python poisson_model.py --help

Nonlinear Model

python nonlinear.py --help

Citation

If you find the work useful for your research, please consider citing the following paper:

@article{celikkanat2024revisiting,
  title={Revisiting K-mer Profile for Effective and Scalable Genome Representation Learning},
  author={Celikkanat, Abdulkadir and Masegosa, Andres R. and Nielsen, Thomas D.},
  journal={Advances in Neural Information Processing Systems},
  volume={37},
  year={2024}
}