Single cell RNA-seq anlaysis

Question

[kanwarjag]

There are several posts about single cell RNA-seq. However, I want to understand which may be the best workflow in terms of sensitivity and specificity as well generating some good graphics/ matrix for figures.

Thanks

Answer 1

[rpolicastro]

Since this post has accumulated views, I'll add a few resources.

I would consider the three most popular scRNA-seq frameworks to be Seurat (R), Bioconductor (R), and Scanpy (Python).

I'll speak in generalities, but Seurat is likely the most popular of the frameworks, and tends to introduce features before the other workflows. Bioconductor I like because they use common data structures, making it more extensible and easier to work with than Seurat. I also find the documentation to be the best of the three options. Scanpy has most of what Seurat and Bioconductor have, just implemented in Python.

For trajectory analysis Monocle 3, slingshot, and tradeSeq are written in R, so will be easier to work with if using Seurat or Bioconductor. PAGA and scVelo (for RNA-velocity) are written in Python and are seamlessly integrated with Scanpy, since they are by the same lab. There are ways to convert all data structures to any other type of data structure (with varying degrees of ease), so don't let this limit you.

Answer 2

[jared.andrews07]

To add to rpolicastro's excellent answer, Bioconductor has the added benefit of the Orchestrating Single-Cell Analysis book, which is an unparalleled resource in my mind. Even if you don't use Bioconductor packages for your analysis, it does a great job explaining the why behind many steps of single-cell analysis that most other packages/documentation tend to gloss over. It also has a wide array of examples and lots of code you can crib.

For viz, I am biased since I have been involved in its development, but dittoSeq is my go-to package, as it works natively with Seurat, SingleCellExperiment, and SummarizedExperiment data structures. This means it is useful for both bulk and single-cell RNA-seq studies and can work with DEseq2/edgeR output very easily as well. It's also color-blind friendly by default, which is particularly problematic for single-cell RNA-seq images.

velociraptor is also a new Bioconductor package that wraps the scVelo method that Rob mentioned, which was previously a bit of a pain to use with SingleCellExperiment objects that had been run through Bioconductor-based analyses. Scanpy and PAGA are now much more interoperable with the Bioconductor ecosystem since the release of the zellkonverter package as well.