nf-core / deepmodeloptim

Stochastic Testing and Input Manipulation for Unbiased Learning Systems
https://nf-co.re/deepmodeloptim
MIT License
23 stars 9 forks source link

[future][discussion] how to incorporate external modules ? #195

Open mathysgrapotte opened 1 month ago

mathysgrapotte commented 1 month ago

Description of feature

Allow for data processing to be done by modules external from the stimulus-py package and not necessarily in python

Example :

Splitting a sequence dataset could be done using pairwise similarity + kmeans There are many tools doing pairwise similarity, some even gpu accelerated, and some nf-core modules could do this

Current way to do this would be to wrap the method in python and bumb the pip package but it isn't the nf-core way, it would be much better if instead, we could re-use modules with some minor overhead for format/data processing.

This should be done while considering :

  1. whether the tool should be ran or not depends on the experiment config
  2. code should be kept clean (there are many tools that could interface with data processing - and the number will keep growing ), chaining IF/ELSE is not an option
  3. good error handling (i.e. blast can't run on images for instance)