miRBind is a machine learning method based on ResNet. It learns the rules of miRNA:target binding and provides a probability for the potential binding of a pair of given miRNA and target sequence.
Have a look at our paper miRBind: A Deep Learning Method for miRNA Binding Classification for more information about our work.
The user-friendly miRBind web application for performing predictions https://ml-bioinfo-ceitec.github.io/miRBind/
Using Git:
git clone https://github.com/ML-Bioinfo-CEITEC/miRBind.git
git clone git@gitlab.com:RBP_Bioinformatics/miRBind.git
mRBind is implemented in python using Keras and Tensorflow backend.
Required:
- python, recommended version 3.7
- Keras 2.7.0
- tensorflow 2.7.0
- pandas
- numpy
#create a virtual environment:
python -m venv venv
#activate it and install the necessary libraries.
source venv/bin/activate
pip install -r requirements.txt
Required input is a tsv file with multiple potential miRNA - target pairs consisting of first column containing miRNA sequence (20 bp long) and second column containing target sequence (50 bp long). To run the model:
cd path/to/miRBind/directory
chmod +x mirbind.py
#if you are not actively sourcing from the previously created virtualenv:
source venv/bin/activate
#run the prediction
./mirbind.py --input <input_file> --output <output_file>
CEITEC MU, RBP Bioinformatics - Panagiotis Alexiou, https://www.ceitec.eu/rbp-bioinformatics-panagiotis-alexiou/rg281