Causal Network Models of SARS-CoV-2 Expression and Aging to Identify Candidates for Drug Repurposing

This repository contains code for Causal Network Models of SARS-CoV-2 Expression and Aging to Identify Candidates for Drug Repurposing (https://arxiv.org/pdf/2006.03735.pdf) main methodology and analysis.

Code_autoencoding directory contains code for training an over-parameterized autoencoder for mining relevant drugs for drug repurposing. Code_ppi directory contains scripts and Jupyter notebooks for running Steiner tree and causal analysis for investigating the drug mechanism and prioritizing drugs.

System requirements

Ubuntu 16.04, Python 3.7, PyTorch 1.6 (CUDA enabled), cmapPy 4.0.1, [OmicsIntegrator 2](omicsintegrator 2.3.10, https://github.com/fraenkel-lab/OmicsIntegrator2), causaldag 0.1a133, networkx (2.4, note that omicsintegrator 2.3.10 requires networkx==2.1), scikit-learn (0.22.2), graphviz (2.40.1)

Installation guide

Clone this repository with the code (~5 secs):

git clone http://github.com/uhlerlab/covid19_repurposing.git

Demo

For demos of autoencoder and protein-protein interaction analysis, see the READMEs in Code_autoencoding and Code_ppi, respectively. The expected outputs are shown in the accompanying Jupyter notebooks in Code_ppi/Code/SteinerTree_notebook.ipynb and Code_ppi/Code/CausalAnalysis.ipynb. The expected runtime of the Jupyter notebooks is < 2 hours and the autoencoder training is < 2 hours.

Instructions for use

In order to run the software on your data replace inputs in Jupyter notebooks in Code_ppi and scripts in Code_autoencoding with your own data.