DeepSVR

This repository can be used to recapitulate the development and analysis of a machine learning model approach to somatic variant refinement. DeepSVR contains the raw data (e.g. bam files and manual review labels), code required for data preparation, and validation sets to test the ultimate models. Using the prepared data, we developed three machine learning models (Logistic Regression, Random Forest, and feed-forward Deep Learning). The model that was most consistent with manual revew labels was the Deep Learning model. This model was packaged and is available for use.

A walk-through of the DeepSVR repo can be found on the Wiki page.

Installation of deepsvr package

Note: Please ensure that you are running these commands using python3 or greater.

1) Clone the DeepSVR GitHub Repo see Repository - Installation

2) Install Anaconda see Downloads - Anaconda

3) Add BioConda Channels

conda config --add channels defaults
conda config --add channels conda-forge
conda config --add channels bioconda

4) Install DeepSVR see BioConda - DeepSVR

conda install deepsvr

5) Test installation and view DeepSVR options

deepsvr --help

Using deepsvr with docker

1) Clone the DeepSVR GitHub Repo see Repository - Installation

2) Build docker image

docker build -t deepsvr .

3) Test installation and view DeepSVR options

docker run deepsvr --help

4) Using repo data inside the docker container

docker run -v `pwd`:/code deepsvr

Data passed to deepsvr tool needs to be available inside the container. So binding the repo directory path to /code inside the container allows to access for example training data as /code/wiki_figures/create_classifier/training_data_call.pkl.