synergy-screen

July 18, 2025 · View on GitHub

⚠️ WARNING:
The new URL for the synergy-screen app is https://synergy-screen.netlify.app/.
Please update your bookmarks

See README_STATIC_CONVERSION.md for details about the static site conversion process.

Purpose

This git repo contains the companion code the manuscript Statistical Assessment and Visualization of Synergies for Large-Scale Sparse Drug Combination Datasets, by Arnaud Amzallag, Sridhar Ramaswamy, Cyril H. Benes, from the MGH Cancer Center.

This script allows to reproduce, from raw data, the statistical analysis for the significance assessment of synergies in the systematic drug screen performed at the MGH Cancer Center. The script is thoroughly commented and is intended to be read and executed on the R prompt. It produces several diagnostic and explainatory figures. this code should:

  • provide a transparent description of the methods used in the study
  • expose a novel way of estimating single drugs viability for estimation of drug synergy
  • facilitate follow-up analysis and exploration of the large combination data published with the study

In addition, it provide the code of a web app for rapid interactive visualization of the 439,128 drug combination viabilities provided, which is available at https://synergy-screen.netlify.app/

NEW: This repository also includes a static Hugo website version that provides the same visualization capabilities without requiring a Node.js server. The static site can be deployed to any web hosting service.

Usage

The static Hugo website provides the same drug combination visualization without requiring a running server:

  1. Run data preprocessing (required before building the Hugo site):

    cd static_preprocessing/
./preprocess_modern.sh

  2. Build and serve the Hugo site:

    cd ../hugo-site/
# For development with live reload:
hugo server
# Or build static files for deployment:
hugo

  3. Access the site:

    • Development: http://localhost:1313
    • Production: Deploy the public/ folder to any web hosting service

Important: You must run the preprocessing script (preprocess_modern.sh) before starting Hugo, as it generates the required data files that the Hugo site depends on.

Pre-requisites

We recommand running the R analysis on a machine with at least 8 Gb of memory.

The following packages must be installed: dplyr, reshape2, readr and ggplot2.

Processing in R

To understand the processing and analysis steps from raw data, we recommand to follow the commented script combos_script.R and execute it line by line. However, if you just want to obtain all the output files at once (including the necessary files to run the web app), you can run the full script with

   Rscript combos_script.R

on the command line. It should take between 5 and 10 minutes to run.

Running the legacy web application (Node.js server)

Note: This is the original dynamic web application. For most use cases, the static Hugo website (above) is recommended.

  • Run combos_script.R (see above) to produce the summarized data files that are used for the web visualization.
  • Install npm, since the web app is running on Node.js, a lightweight web server written in javascript.
  • From the directory combo_web, run npm install (you might need sudo npm install, on ubuntu for instance).
  • Run node server.js
  • In a web browser, go to http://localhost:3000/combo_app.html

Running with Docker

For convenience, a docker image is available on dockerhub.org, and the current web application can be started with the following command:

    docker run --rm -p 3000:3000 -d arnaudmgh/synergy-screen-app

The Dockerfile is available in this repository.

Running with Docker on AWS

Additionally, the script aws-init.sh can be used as User Data to start an AWS instance (from a standard Amazon AMI) to host the web application in a private AWS cloud.

  • The security group should allow TCP port 3000 inbound from anywhere;
  • The web application will be accessible at http://<your-instance-IP>:3000/combo_app.html

Troubleshooting

  • node server.js returns Error: ENOENT: no such file or directory, open './combo_all_combos2.csv'. Have you run the R script combos_script.R, and did it write the file combo_all_combos2.csv in the right directory (combo_web)?
  • The script does not finish: It could be that your machine does not have enough memory. We recommend at least 8 Gb of memory.