Scorpio

Welcome to the Scorpio project! This repository contains advanced tools for training triplet networks using contrastive learning on diverse DNA sequences and data from promoter detection, phylogenomic analysis, antimicrobial resistance (AMR) detection, and any hierarchical information, which can improve downstream analysis and insights.

Tutorials

The GitHub Wiki also contains tutorials to help you learn how to use Scorpio tools with real data.

Training Data

For training the gene-taxa model with full genes, we have included the data in this Zenodo record: Zenodo. Please follow the instructions below to fully download the data. This data can be used with the trainer to train and save the model:

wget https://zenodo.org/api/records/12175913/files-archive -O scorpio-gene-taxa.zip
unzip scorpio-gene-taxa.zip -d scorpio-gene-taxa

Models Available at Zenodo

Scorpio Models

• ScorpioBigDynamic
  https://zenodo.org/record/14176840

• ScorpioBigEmbed
  https://zenodo.org/records/14176823

BigBird Model

Our pre-trained model, MetaBERTa, a version of BigBird trained on gene sequences, is available here:
MetaBERTa-BigBird-Gene on Hugging Face

Installation

You can set up the environment for scorpio using either a conda environment or a Docker image. Follow the instructions below for your preferred method:

Method 1: Conda Environment

1. Create a conda environment named scorpio based on the environment file in the src directory:

   conda env create -f src/environment.yml -n scorpio

2. Activate the conda environment:

   conda activate scorpio

3. Run the setup script to add scorpio to your PATH:

   ./src/setup.sh

Method 2: Docker

1. Download and run the Docker image:

   docker pull eesilab/scorpio
   docker run -it eesilab/scorpio

After following the steps for either method, your environment should be set up and ready to use the scorpio tool.

Get Involved

We encourage community involvement and welcome contributions to the Scorpio project (Report Issues, Submit Pull Requests, Join Discussions).

Please email us for any inquiries.

Maintainer: Saleh Refahi (sr3622 at drexel dot edu)

Owner: Gail Rosen (gailr26 at drexel dot edu)

Citation :

@article{refahi2025enhancing,
  title={Enhancing nucleotide sequence representations in genomic analysis with contrastive optimization},
  author={Refahi, Mohammadsaleh and Sokhansanj, Bahrad A and Mell, Joshua C and Brown, James R and Yoo, Hyunwoo and Hearne, Gavin and Rosen, Gail L},
  journal={Communications Biology},
  volume={8},
  number={1},
  pages={517},
  year={2025},
  publisher={Nature Publishing Group UK London}
}