Sample data

[mccalluc]

@ngehlenborg: No hurry,

• but can you point me at a dataset that would be good to used for a first demo? Or suggest how many rows and columns should be accommodated?
• For the volcano plot and the pca, is it worthwhile to do it in JS? It would make it easier to plug in new sample data, and maybe it could be useful for smaller datasets in the long run, too.

[ngehlenborg]

Still looking for a good sample data set.

Typical data set sizes we need to be prepared for: ~25,000 genes (= rows) and say 100 conditions/samples (= columns).

In heatmaps, the genes will be filtered to something more managable (dozens to hundreds) and obivously we can't render 25000 rows even if we make them 1 pixel high. There will either be overplotting (initial solution) or we find smart ways to aggregate (later solution?).

The scatterplots (PCA and volcano) should support the same number of items as the worst case (~25,000), which is best handled using opacity (which will make rendering much slower). Note that PCA can be done on both the genes and the samples (which is much easier to handle in the visualization). I suggest you build on the canvas-based scatter plot that @sgratzl started to work on.