Multi-level factors and continuous covariates

[PauBadiaM]

Hi devs,

Amazing work! It's really nice to have such a strong DEA framework finally available in python. Here's a suggestion regarding factors:

Is your feature request related to a problem? Please describe.
Currently pydeseq2 is limited to bi-level factors, limiting its usage whenever there are multi-level factors or continuous covariates.

Describe the solution you'd like
I know it might be cumbersome to implement, but it would be convenient to add support to multi-level factors and continuous covariates in the design_factors argument of DeseqDataSet.

Describe alternatives you've considered
For multi-level factors, the current alternative is to generate a single DeseqDataSet for each combination of groups, making it tedious to deal with (although it's true that depending on the within-group variability differences it might be needed as explained in the original docs).

For continuous covariates, the current alternative is to cut them into bins and generate factors of them, which results in a multi-level factor with the previously mentioned problems.

[BorisMuzellec]

Hi @PauBadiaM, thanks for your interest in PyDESeq2!

I'm currently working on multi-level designs (see PR #101), hopefully they should be supported in a couple of days.

Unfortunately, we probably won't have the bandwidth to work on continuous variables in the near future, but we'd be glad to review a PR should someone want to implement continuous covariates :) !

[laserson]

+1 for continuous covariates!

[mbernste]

Really great work.

+1 for continuous covariates. At a high level, what changes/refactoring would be needed to support this?

[mikecuoco]

+1 for continuous covariates! So useful for most experimental designs

[njbernstein]

Adding continuous variables would be wonderful

[myz540]

+1 for continuous covariates