plotomics for R • plotomics

Fifteen GPU- and canvas-accelerated visualization widgets for bioinformatics, built on a shared JavaScript core and exposed to R through htmlwidgets. Every widget renders in the RStudio Viewer, R Markdown, Quarto and Shiny, and each ships a matching *Output() / render*() pair for Shiny apps.

The same core drives the Python and JavaScript packages, so a widget looks and behaves identically wherever it ships.

Installation

install.packages("plotomics")

For the development version, from r-universe or GitHub:

install.packages("plotomics", repos = "https://samuelbharti.r-universe.dev")

# install.packages("pak")
pak::pak("samuelbharti/plotomics/pkg-r")

Quick start

library(plotomics)

# Differential expression
volcano(data.frame(
  x    = res$log2FoldChange,
  y    = -log10(res$padj),
  gene = rownames(res)
))

# A single-cell embedding: a factor pins the legend order and keeps
# unused levels, the way drop = FALSE does in ggplot2
embedding(data.frame(
  x     = umap[, 1],
  y     = umap[, 2],
  color = factor(cell_type)
))

# Kaplan-Meier, straight from a survfit object
km(survival::survfit(survival::Surv(time, status) ~ sex, data = lung))

Components

Area	Functions
Expression and abundance	`volcano()`, `bioheatmap()`, `clustermap()`, `dotplot()`, `violin()`
Single-cell and spatial	`embedding()`, `spatial()`
Cohort and variant	`oncoplot()`, `lollipop()`, `km()`, `bioprofile()`
Sets, hierarchies, networks	`upset()`, `treemap()`, `network()`
Genome and chromatin	`hic()`

The Python and JavaScript packages ship two more: igv() and gosling(), both genome browsers. They came out of the R package before the CRAN submission, because their bundles push the installed size past CRAN’s 5 MB limit. Whether they return is open.

Helpers: oncoplot_memo_sort() for the conventional oncoplot column order, upset_intersections() for exclusive set intersections, and violin_density() for densities computed in R.

Two names differ from the obvious choice, so that attaching the package masks nothing in base or the recommended packages: bioheatmap() rather than heatmap(), and bioprofile() rather than profile(). Both have *_plotomics() aliases (heatmap_plotomics(), profile_plotomics()).

Shiny

Every widget has a Shiny pair. The network and embedding widgets also report selections back to the server:

ui <- fluidPage(networkOutput("net"))

server <- function(input, output) {
  output$net <- renderNetwork(network(nodes, edges))
  # clicking a node sets input$net_selected
  observeEvent(input$net_selected, print(input$net_selected))
}

Built for large data

Numeric columns reach the browser as a binary buffer rather than JSON, which is what keeps several hundred thousand points interactive rather than merely drawable.

Documentation

Reference for every function
Changelog
Project overview: why the package exists, the Python and JavaScript packages, and when an established package such as ComplexHeatmap or survminer is the better choice

License

MIT. See LICENSE.