Abstracted definition of a module

[zouter]

type: dataset_generator
design@R: |
  crossing(
    differentially_expressed_rate = seq(0.1, 0.9, 0.2)
  ) %>% 
    mutate(
      num_cells = runif(n(), 10, 1000),
      seed = row_number(),
      id = as.character(row_number())
    )
output:
  - expression
  - tde_overall
  - meta

[zouter]

How to handle different inputs and outputs though? For a metric module:

• Dataset - model
• Model
• Model 1 - Model 2

[zouter]

Also how to handle multiple designs in one module? Eg. see komparo/tde_method_controls

[zouter]

We need some kind of module blueprint, which is by default free, but which can be restricted eg. for dataset generators, methods, metrics, and report generators