aarthir239/deepregfinder repository overview

⁠DeepRegFinder: Deep Learning based Regulatory Elements Finder

By Aarthi Ramakrishnan, George Wangensteen, Sarah Kim and Li Shen

Icahn School of Medicine at Mount Sinai, New York, NY, USA

DeepRegFinder is a deep learning based program to identify DNA regulatory elements using ChIP-seq. It uses the deep learning framework PyTorch.

DeepRegFinder @ GitHub : https://github.com/shenlab-sinai/DeepRegFinder⁠

Paper : https://www.biorxiv.org/content/10.1101/2021.04.27.441658v1⁠

⁠Overview

Identifying DNA regulatory elements such as enhancers and promoters has always been an important topic in the epigenomics field. Although certain histone marks are known to exhibit characteristic binding patterns at enhancers and promoters, the exact rules to make the call do not exist. Using machine learning models that are trained on known enhancers to predict at other regions using histone mark ChIP-seq data has been found to be the most successful method so far. Many machine learning algorithms for enhancer identification exist. However, most of them are designed for reproducing results only. It's a hassle to apply them to your own data considering the most time-consuming part of a machine learning project is often data cleaning and formatting. We developed DeepRegFinder to be a modularized pipeline for you to build training data from aligned reads and genomic annotation easily so that you can use them to train models and make predictions. DeepRegFinder uses two deep neural networks: convolutional neural net (CNN) and recurrent neural net (RNN).

⁠How to run this image?

Assuming your workstation has Docker⁠ installed, open a terminal window and run -

docker pull aarthir239/deepregfinder

Once the docker image is pulled, it is ready to be used! Following are the commands that may be executed on your workstation using an example dataset at this link - https://drive.google.com/drive/folders/1sW9KM9TnK6nqquf7nQniEpfTtiKtWVni⁠

⁠Preprocessing -

docker run -v /local/path/to/example_dat/:/example_dat \
			aarthir239/deepregfinder \
			drfinder-preprocessing.py /example_dat/preprocessing_data.yaml /example_dat/output

⁠Training -

docker run -v /local/path/to/example_dat/:/example_dat \
		aarthir239/deepregfinder \
		drfinder-training.py /example_dat/training_data.yaml /example_dat/output

⁠Prediction -

docker run -v /local/path/to/example_dat/:/example_dat \
			aarthir239/deepregfinder \
			drfinder-prediction.py /example_dat/wg_prediction_data.yaml /example_dat/output