Home
Framework for Interpretable Neural Networks for genetics
- What is GenNet?
- Getting started
- GenNet from command line.
- GenNet.py convert
- GenNet.py train
- GenNet.py plot
GenNet is a command line tool that can be used to create neural networks for (mainly) genetics. GenNet gives the opportunity to let you decide what should be connected to what. Any information that groups knowledge can therefore be used to define connections in the network. For example, gene annotations can be used to group genetic variants into genes, as seen in the first layer of the image. This creates meaningful and interpretable connections. When the network is trained the network learns which connections are important for the predicted phenotype and assigns these connections a higher weight. For more information about the framework and the interpretation read the
The Gennet framework is based on tensorflow, click here for the custom layer.
-
GenNet uses CUDA. Please make sure you have the correct version of CUDA installed. GenNet has been tested for:
- CUDA 9.1 & Tensorflow 1.12.0
- CUDA 10.0 & Tensorflow 1.13.1
- CUDA 10.0 & Tensorflow 2.0.0-beta1
- CUDA 10.1 & Tensorflow 2.2.0
Open terminal. Navigate to the a place where you want to store the project. Clone the repository:
git clone https://github.com/arnovanhilten/GenNet
Navigate to the home folder and create a virtual environment
cd ~
python3 -m venv env_GenNet
This automatically instals the latest Tensorflow version for which GenNet has been tested. If you have an older version of CUDA install the appriopriate tensorflow-gpu by
pip install tensorflow-gpu==1.13.1 (change 1.13.1 to your version).
Activate the environment
source env_GenNet/bin/activate
Install the packages
pip3 install --upgrade pip
pip3 install -r requirements_GenNet.txt
Gennet is ready to use!
The command l
Open the command line and navigate to the GenNet folder
start the command line program by:
python GenNet.py
type --help for help
convert bed and vcf to .h5
Training the network
Generate plots