Operational dialogue

[grpinto]

So, I essentially made DIALOGUE work with the toy example and without confounders which is not available in the original implementation. Things that still need to be done :

• Allow multiple confounders, currently we can only handle the original cellQ and none
• Improve the input output structure, currently it only creates a big dictionary similar to what is done in the R version, but I think this can be streamlined very easily by using an anndata
• Improve the performance time for real datasets (dialogue toy example runs fast). Major chokepoints is in the stats model regression in multilevel modelling
• We also need to implement one additional function in Dialogue called pheno which seems to me to be a summary of the results

** Very important **

To run Dialogue notebooks you can get the MCPs simply by accessing to the "residual_scores" key of the output dictionary of DIALOGUE1 I have tried it myself and it works fine. I changed the Test association function so it can run with the result from calculate_multifactor_PMD with the kay "residual_scores" you just need to provide it to the function

Changes performed :

• Defined a lot of new functions, they are too many to refer here, but I think the code works fine. Can do if really necessary