DataMap: A Browser-based App for Visualizing High-Dimensional Data

1. Introduction

High-dimensional datasets, such as expression matrices from RNA-seq or proteomics experiments, have become commonplace in biomedical research. Heatmaps are an effective visualization method, efficiently representing thousands of data points in a matrix through variations in color. Several heatmap-centric visualization tools, including Clustergrammer¹, Phantasus², and Morpheus³, have been developed to facilitate the analysis of these expansive datasets. Phantasus and Morpheus operate entirely within the user’s web browser, while Clustergrammer requires server-side processing.

Our goal was to develop a secure, browser-based application capable of generating high-quality and reproducible visualizations. To address this, we developed DataMap, an R/Shiny application deployed via Shinylive, leveraging WebR—a specialized version of R compiled into WebAssembly—to enable execution directly within web browsers. Hosted as static files on GitHub, this serverless design ensures sensitive data remains securely on the user’s device while eliminating the need for server-side computation.

DataMap supports multiple visualization methods, including hierarchical clustering with heatmaps, principal component analysis (PCA), and t-distributed stochastic neighbor embedding (t-SNE)⁴. These methods enable researchers to uncover biologically significant patterns, clusters, and relationships within complex datasets. To facilitate usability, the application automatically recommends optimal file-parsing methods and data transformation settings by analyzing input files and assessing data distributions. Although optimized for omics datasets, DataMap remains broadly applicable to general high-dimensional data matrices across various research domains.

2. Implementation

DataMap is implemented as an R/Shiny application compiled into WebAssembly via Shinylive, enabling entirely client-side execution within web browsers. The application is hosted on GitHub Pages as static files, automatically updated through GitHub Actions from the source code repository, ensuring continuous integration and continuous deployment (CI/CD).

The app consists of several functional modules:

This modular and serverless design ensures data security by processing exclusively on the user’s device and facilitates easy maintenance and ongoing enhancements through automated deployment.

3. Operation

To use DataMap, simply access the application via the static GitHub page at https://gexijin.github.io/datamap using modern web browsers such as Chrome, Edge, Safari, or Firefox. Additionally, users may install and run DataMap locally as an R package using the following commands in R:

This flexibility ensures seamless operation both online and offline, accommodating diverse user needs.

4. Results

4.2 Comparison with existing tools

DataMap complements existing visualization tools such as Clustergrammer, Phantasus, and Morpheus. Like Phantasus and Morpheus, DataMap employs client-side processing for enhanced data security. It extends their functionality by offering a broader set of preprocessing options, automatic generation of reproducible R scripts, and publication-quality graphics. However, DataMap is less interactive than native web applications built with Java or other programming languages.

Use cases

We used DataMap to visualize genes upregulated by ionizing radiation in mouse B cells with and without functional p53 gene⁶. Figure 1A shows the top genes specifically induced in B cells with p53. Experimental conditions (genotype and radiation exposure) are annotated by column annotations, while row color bars indicate genes involved in apoptosis. This heatmap clearly reveals genes strongly responsive to radiation only in wild-type B cells, highlighting the functional importance of p53. In Figure 1B, we visualize a t-SNE projection of 2,700 single-cell RNA-seq samples, color-coded by clusters corresponding to cell types. These examples highlight DataMap’s capability to uncover insights within complex, high-dimensional omics datasets.

Figure 1. Example visualizations.

(A) Top 20 genes upregulated by ionizing radiation in mouse B cells with functional p53 gene⁶, and (B) t-SNE projection of 2700 single-cell RNAseq profiles of peripheral blood mononuclear cells (PBMCs), available from 10X Genomics. Both datasets are included as built-in examples within the application.

5. Discussion

When analyzing large datasets, browser-based execution is slower compared to native execution. For example, generating a hierarchical clustering heatmap of a 2700×50 matrix takes approximately 80 seconds when run in the browser, compared to just 5 seconds in native R on the same laptop (Intel 11th Gen Core i7-1185G7, 3.00 GHz). Users are encouraged to install DataMap locally as an R package for extremely large datasets. Another limitation stems from DataMap’s reliance on the WebR, which only supports a subset of R packages with delayed updates.

In summary, DataMap combines secure client-side processing with robust data preprocessing and reproducible workflows. It complements existing web-based tools, equipping biomedical researchers with a powerful tool for exploratory analysis. Future development will focus on expanding visualization capabilities and incorporating additional analytical modules.

Software availability

Software available from: https://gexijin.github.io/datamap

Source code available from: https://github.com/gexijin/datamap

Archived source code at time of publication: https://doi.org/10.5281/zenodo.15361414

License: GNU General Public License v3.0