Interactive plotting

[grst]

The suggestion was brought up to build an interactive version of the vdj_usage sankey plot, to appreciate less abundant gene combinations by hovering over the respective segments.

While this single interactive plot could probably be straightforwardly solved implementing either

• the plotly example or
• the vega example,

I would like to first discuss interactive visualization in a broader scope.

Considerations

• Ideally use one single library throughout all visualizations to avoid the "dependency hell"
• In any case, it should be an optional dependency (i.e. be imported upon function call only)
• There is a similar discussion in Scanpy (https://github.com/theislab/scanpy/issues/253) and some demo examples: https://github.com/theislab/interactive_plotting. (implemented using Bokeh and Holoview it seems)

Interactive plotting libraries (non-exhaustive)

• Plotly
• vega
• bokeh
• mpld3
• Holoview

Plot that would profit a lot from interactivity

• vdj_usage
• clonotype_network
• "linked plots", e.g. selecting cells in UMAP and updating group_abundance automatically.