Interactive Visualization - Usage Guide

Overview

Interactive HTML/web plots support exploration via zoom, pan, hover, and brushing. plotly is the modern Python and R default (ggplotly converts ggplot2 directly); bokeh suits server-side dashboards; gganimate / plotly frames cover animation. The static-export pipeline changed materially in 2025 - orca is EOL; Kaleido v1+ is the standard but no longer bundles Chrome and dropped EPS. Always produce static alongside interactive for journal submission.

Prerequisites

pip install plotly bokeh kaleido altair
# Static export requires installed Chrome/Chromium

install.packages(c('plotly', 'gganimate', 'htmlwidgets', 'DT', 'leaflet'))

Quick Start

Tell your AI agent what you want to do: - "Make this scatter interactive with plotly - hover shows sample ID" - "Convert this ggplot to plotly via ggplotly" - "WebGL acceleration for 20000-point scatter" - "Static PDF export via Kaleido v1; verify Chrome is installed" - "Animate time-course UMAP with plotly frames"

Example Prompts

Interactive volcano

"Build interactive volcano with plotly Express; hover_data = ['gene', 'baseMean', 'padj']. Reference lines at fdr=0.05 and lfc=1. WebGL for >5000 points."

ggplotly conversion

"Take this ggplot and pass to ggplotly with tooltip = c('text', 'x', 'y'). Save as self-contained HTML via htmlwidgets."

Static export

"Export the plotly figure as a 300 DPI PDF for the paper. Use Kaleido v1; verify Chrome is installed; produce a fallback PNG if PDF fails."

Animated UMAP time-course

"Animate UMAP across 5 timepoints using plotly animation_frame='timepoint'. Constant axis range across frames."

Dashboard

"Build a Bokeh server dashboard with linked-brushing scatter + bar chart for cohort exploration."

What the Agent Will Do

Decide interactive vs static: notebook exploration / supplement / dashboard -> interactive; journal figure -> static (alongside).
For Python: plotly Express for fast onboarding; graph_objects for fine control; Scattergl for >5000 points.
For R: prefer ggplotly for ggplot conversion; htmlwidgets ecosystem for tables/maps/networks.
Configure hover_data, tooltips, color/size aesthetics.
For static export: confirm Kaleido v1 + Chrome installed; use fig.write_image() with no engine=.
For animation: limit to ≤100 frames; pre-aggregate per-frame data; constant axis range.
Save HTML as self-contained (htmlwidgets::saveWidget(..., selfcontained = TRUE)) for portability.
Always produce static PDF alongside HTML for journal submission.

Tips

Always produce static alongside interactive. Interactive HTML cannot be cited as a paper figure.
Kaleido v1 is the current static-export engine. No engine= argument needed; orca is EOL.
Kaleido v1 requires installed Chrome/Chromium. No longer bundled. Verify with which chrome or equivalent.
EPS export dropped in Kaleido v1. Export PDF then convert via pdf2ps (ghostscript). For pure-vector EPS, may need a separate path.
WebGL (Scattergl) for >5000 points. Default Scatter is slow at scale.
HTML > 10 MB is a warning sign. Pre-aggregate (Datashader) or filter before exporting.
ggplotly is the lowest-friction R interactive path - write ggplot, get plotly.
Self-contained HTML for portability (selfcontained = TRUE in saveWidget). Default CDN-linked breaks offline.
For animation, cap at ~100 frames. File size and viewer attention both degrade.
bokeh for dashboards / streaming; plotly for static-figure-equivalent interactivity.
Don't cite interactive in figure captions. Treat as supplement; static for the paper.

reporting/quarto-reports - Embed in scientific reports
reporting/rmarkdown-reports - htmlwidgets in Rmd
data-visualization/ggplot2-fundamentals - ggplot for ggplotly conversion
data-visualization/dimensionality-reduction-plots - Interactive UMAP/PCA
data-visualization/network-visualization - PyVis interactive networks