Orca-Fly

The repository contains code for Orca-Fly, an adaptiation of Orca frameworks for multi-scale genome structure prediction in early Drosophila embryogenesis. The manuscript is now published here.

Get started

If you just need predictions for one or a handful of variants, we have provided the core functionalities on a web server.

Install and run Orca-Fly locally

Pre-requisites: Python environment and Selene library.

1. orca_env Python environment

   The enviornment setup involves 3 steps, install (1) Python 3.9, (2) Pytorch, and (3) the remaining packages. This is to save time for conda to solve the dependency for Pytorch.

• Install conda/miniconda if you don't have it already.
• Create the environment from the orca_env_part1.yml file: conda env create -f orca_env_part1.yml
• Activate the environment: conda activate orca_env
• Install Pytorch following the Pytorch installation guide, choose the appropriate parameters (CPU or GPU, OS, etc.) for your system.
• Install the remaining packages: conda env update -f orca_env_part2.yml

2. Install Selene (under the orca_env environment)

git clone https://github.com/kathyxchen/selene.git
cd selene
git checkout custom_target_support
python setup.py build_ext --inplace
python setup.py install 

Now you are ready to run Orca-Fly locally, with the orca repository cloned and resource files downloaded.

Clone the Orca repository

git clone https://github.com/jzhoulab/Orca-Fly.git
cd orca

Predict with pretrained models

Please download the dataset and pretrained model weight by the link [TBD]. You can use predict.ipynb to get predictions of interested regions.

Train Orca-Fly models

If you have set up Orca with its dependencies and has the necessary GPU resources, you can train new models following the example code under the train directory to train new Orca-Fly models.