Improve performances and interface of `tidygate` for large single-cell objects

[stemangiola]

tidygate allows gating points from a tibble, visually procedurally and in a replicable way, compatibly with a pipe-oriented programming style

https://github.com/stemangiola/tidygate

With an arbitrarily large amount of points (1M cells), it struggles, mostly at the visualisation step.

We propose a downsample strategy for the visual preview, while the selection would be done on all data points.

PLEASE IGNORE THIS AT THE MOMENT: As a possible second PR, the plotting backend might be changed from the old interactive plotting style base-R (often broken in the R studio server). A more modern Photoshop, the lasso-style selector, would be great. An example is CellSelector() from Seurat. However, we would need to implement lasso-style gating and multiple gating.

[stemangiola]

@LiNk-NY this is one of the challenges I was mentioning.

[stemangiola]

@william-hutchison

[multimeric]

So, to make this more concrete, you want to propose a downsample method for visual previews? And in addition, you think that there could be a more efficient selection method? Do you mean a spatial data structure that would be well suited to filtering?

[stemangiola]

So, to make this more concrete, you want to propose a downsample method for visual previews?

Yes I updated above.

And in addition, you think that there could be a more efficient selection method? Do you mean a spatial data structure that would be well suited to filtering?

Interesting idea. Not sure about this, this would require some research. But it is out of scope for the moment.