Multimodal Data

I wanted to open a discussion about how we deal with CITEseq dataset so that we can align and if necessary adjust the standard workflow accordingly.

Some relevant sources:
- new tutorial from scanpy people (not properly published yet, so probably not finalized version): https://scanpy-tutorials.readthedocs.io/en/multiomics/cite-seq/pbmc5k.html
- discussion about multimodal data integration in scanpy: https://github.com/theislab/anndata/issues/237
- publication on the normalization of citeseq data (preprint, methodology implemented in R): https://www.biorxiv.org/content/10.1101/2020.02.24.963603v1

So far when I have worked with CITEseq data I did the following:
- I separated protein and gene expression data
- I performed filtering only on the gene expression data and then applied the same filter to the protein expression data
- I normalized gene expression and protein expression data individually (using the same normalize_per_cell methodology)
- I merged the two back together and took the log1p

For clustering it depended on my application if I ran it on only genes, only proteins or a mixture of both.

No repository files, entry points, or tests are named. Start by reviewing the linked Scanpy tutorial, the anndata discussion, and the CITE-seq normalization publication, then compare them with the workflow described in this issue. Done would require an agreed standard workflow for multimodal data and a clearly defined implementation or documentation scope.

Written by the indexing model from the issue text.