Add notebook to demonstrate sweeping cluster parameters with rOpenScPCA

[sjspielman]

Related to #791

This issue tracks adding a notebook in the hello-clusters module that will show how to identify "optimal" clusters with rOpenScPCA::sweep_clusters(). The notebook should explain the concept of why is it important to explore the parameter space, and then provide visualizations with interpretations of which cluster set appears "optimal." The notebook should conclude with adding the chosen clusters into an SCE and/or Seurat object.

[sjspielman]

I've started drafting up this notebook, but before I get too far I thought it would be best to check in about the overall outline for any general feedback! Here's are the sections I'm currently envisioning:

1. Intro section with descriptive text about notebook goals, and link to the 01 notebook as other reference.
2. Introduce sweep_clusters() and calculate by varying a single parameter (nn = seq(10, 30, 10) to keep things small to start)
3. Evaluate clusters:
   • purrr::map() the swept clusters for each of silhouette, purity, and stability
   • for each metric, also visualize and lightly interpret (I'm thinking 1-2 sentences)
   • through i would link to the 01 notebook for more detailed descriptions of individual evaluation functions and the metrics themselves
4. Rinse/repeat in a single section to show varying two parameters at once, since two parameters will have somewhat different plotting code. I intend for this section have much less text and focus primarily (only?) on how you might code it. Accordingly, we might only do one of silhouette or purity here, too.

I'm not necessarily planning to export anything, either, but we might want to show that you can save the full result as rds?

Tagging @jashapiro for thoughts on this outline!

[sjspielman]

@jashapiro just giving this a bump since I'm back in office :)

[jashapiro]

Sorry for not commenting on this sooner. I think your plan looks fine overall. Everything about keeping it small and simple makes sense to me.

[sjspielman]

No worries, thanks!