Skip to content

workflows for big studies #6

Description

@patrickjdanaher

Once studies grow to 20+ tissues, it becomes unreasonable to fit all the data into memory at once. The world needs workflows to handle this.

There are a few categories of solutions:

  1. Some analyses can be run on a subset selected from across the tissues, e.g. a survey of the cell types or niches present.
  2. Run analyses tissue by tissue, e.g. performing cell typing in a consistent way one tissue at a time. (This might call on the results from the subset analysis.) Or, obtain some interesting summary statistic like DE results from each tissue, then report the results across tissues.
  3. Joint analyses of all tissues that load data in batches. E.g., with the right linear algebra, you compute PCs from the whole dataset while only loading one tissue at a time.

It's somewhat complex to set up all the above; I think analysts out there would benefit a lot from some vignettes.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    Status
    Todo

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions