Create a getting started with scpca data section

[allyhawkins]

Closes #52.

Here I'm adding the section on how to get started using an Scpca dataset. I've included sections on importing the RDS files into R, quality control, normalization, downstream analysis, and then working with Seurat.

In the first section I included a short introduction on the two RDS files that are present for each sample and linked out to other sections in the docs when necessary. We do have a FAQ that explains reading into R, but I remembered we talked about repeating some of that information and it seemed like a necessary first step.

For the quality control section, I included information about using miQC and how we have presence of the miQC_pass column in the objects. I have some explanation on what miQC is along with links to the vignette and documentation. I also inserted code chunks to show how you would filter using that column.

In addition to using miQC, we also have stored the basic metrics in the SCE's that we provide and that's what is recommended in OSCA. I included a section on choosing hard cutoffs and tried to caution people to explore diagnostic plots before simply choosing cutoffs. That's why I didn't provide any recommendations on which cutoffs to use there. Some questions I had about the quality control section were:

• Does this information seem like too much detail or too little detail for filtering?
• Should I not include miQC or hard cutoffs?
• Should I include recommended values for hard cutoffs?

For the normalization and feature selection I generally tried to describe what we would recommend while also linking out to other resources so that they could learn on their own outside of our recommendations. I was back and forth on including code chunks in these sections since I don't want people to blindly follow this as a tutorial, but want people to actually read and stay informed. However, the code chunks seemed useful to have to me so they could follow along for each step? Do others have thoughts on this?

In terms of the dimensionality reduction section, I included some detail but did not go too far into the weeds of how PCA and UMAP are calculated. I thought that didn't seem too necessary, but let me know if there should be more detail here.

The last section I included was about using ScPCA data with Seurat. We already have a fairly extensive FAQ on transferring SCE's to a Seurat object so I provided a link to that FAQ followed by tutorials on working with Seurat performing the same steps that I described for working with SCE objects. Should there be more explanation here or more code chunks?

Some other overall questions:

• Does the organization make sense? Would you divide it into different sections?
• Is the level of detail and the amount of linking to external resources what you had in mind?

[allyhawkins]

I went through and incorporated the suggestions from both @jaclyn-taroni and @jashapiro and made the following changes:

• I added the columns description for the colData (taken from the sce contents section), stating that these columns are added by scuttle::addPerCellQCMetrics and can be found in the colData. @jashapiro is this what you had in mind for your comment about adding the names of the columns in?
• I removed the background information in the normalization section, simply stating that the counts were unnormalized and needed to be normalized, provided the code to do so and then some links to more information
• I removed the section on variable features
• I simplified the dimensionality reduction section to have less intro about PCA and UMAP and more here is how to do them. However, I did keep in the cautionary statements about interpreting UMAP since Josh had mentioned he liked having them there. In this section I added a note about selecting variable features and linked to the chapter in OSCA. Is this more what you had in mind here or do you want me to remove more description?
• I added a section on scanpy. I tried to mirror the section on Seurat that I had and link out to the FAQ and then provide a few links. If there are other links that you think would be helpful please let me know as the only reliable ones I had were from the scanpy tutorials and there isn't much there.

[allyhawkins]

I might reduce the UMAP caution section to just the crux, something like: "Use caution when interpreting UMAP results; the location, spacing, and density of the clusters can be dependent on parameter choices and random effects, and may not accurately reflect relationships among cells." I mostly want to get it down to 1-2 sentences, as a reminder of things they have hopefully heard before.

I went ahead and incorporated this suggestion and paired it down to 1 sentence along with incorporating the other minor suggestions. Let me know if this looks as expected.