Network Explorer

Network Explorer enables users to explore single-sample analysis results, capturing relationships between either samples or target sets and recapitulating known disease and biological groupings, respectively.

Interactive network exploration for node-by-observation datasets using an R Markdown Shiny app.

The app reads a structured Excel workbook, computes either:

• a correlation network between nodes across shared observations, or
• a bipartite network linking nodes to observations

It then renders the result as an interactive graph with module detection, edge filtering, node search, and basic network statistics.

Repository Contents

• NetworkExplorer.Rmd: main Shiny R Markdown application
• NetworkExplorerTemplate.xlsx: template workbook for user-supplied input data
• NetworkExplorerMelanoma.xlsx: example workbook
• NetworkExplorerFKTN.xlsx: additional example workbook
• NetworkExplorerLDR.xlsx: additional example workbook
• MIPE_correlations_melanoma.csv: source/example data file included in the repository

Requirements

Install R and the packages used by the app:

install.packages(c(
  "dplyr",
  "igraph",
  "knitr",
  "RColorBrewer",
  "readxl",
  "rmarkdown",
  "shiny",
  "shinydashboard",
  "tibble",
  "tidyr",
  "visNetwork"
))

Running the App

Open NetworkExplorer.Rmd in RStudio and run the document as a Shiny document, or launch it from R with:

rmarkdown::run("NetworkExplorer.Rmd")

Open the app and upload an Excel workbook with the Network data manifest control.

To pre-load a manifest at startup instead of requiring an upload, set NETWORK_EXPLORER_MANIFEST before launching:

NETWORK_EXPLORER_MANIFEST=NetworkExplorerMelanoma.xlsx Rscript run_app.R

Running with Docker

Build the image from the repository root:

docker build -t network-explorer .

If your network requires a different CRAN mirror, pass it at build time:

docker build --build-arg CRAN_REPO=http://cran.r-project.org -t network-explorer .

Run the Shiny R Markdown app on http://localhost:3838:

docker run --rm -p 3838:3838 network-explorer

The running app accepts a user-uploaded manifest from the browser. To pre-load a manifest that already exists in the image, set NETWORK_EXPLORER_MANIFEST:

docker run --rm -p 3838:3838 -e NETWORK_EXPLORER_MANIFEST=NetworkExplorerBiallelic.xlsx network-explorer

To run a different R Markdown file copied into the image, set RMD_FILE:

docker run --rm -p 3838:3838 -e RMD_FILE=NetworkExplorer.Rmd network-explorer

The initial settings for manifest interpretation can be set with environment variables:

NETWORK_EXPLORER_BIPARTITE=true
NETWORK_EXPLORER_USER_PALETTE=false

Input Workbook Format

The app expects an Excel workbook with three sheets:

1. Metadata Contains node-level display metadata. Required columns:

   • Node_Name: The node level objects that will appear in correlation networks. Observations are used to calculate correlations Optional columns:
   • Node_Label: Display name for node in network
   • Color_Factor: Allows custom color designation. Can either be group designations or hexadecimals for more control over palette (specify user_palette = TRUE in script in this case). If left blank, nodes are automatically colored by louvain cluster membership
   • Shape_Factor: Similar to color, only accepts group designations at this time

2. Data sheet Long-format measurements with columns:

   • Node
   • Observation
   • Value

3. Data_Mask Used to exclude nodes or observations before the network is built.

   • columns 1-2 for node masking
   • columns 4-5 for observation masking

Use NetworkExplorerTemplate.xlsx as the starting point for new datasets.

Features

• Correlation network generation from node-by-observation values
• Optional bipartite graph construction
• Louvain community detection
• Correlation cutoff filtering
• Positive-only or negative-only edge filtering
• Removal of singleton nodes
• Exact-match node search and focus
• Node label, color, and shape customization from metadata

License

Network Explorer is licensed under the Creative Commons Attribution 4.0 International License. When using or adapting this work, please attribute it as:

Network Explorer, National Center for Advancing Translational Sciences (NCATS), https://github.com/ncats/NetworkExplorer

Include a link to the license and indicate whether changes were made.

Contact

Andrew Patt Ewy Mathé