add plotMA() functionality

[emmcauley]

Reference Issue or PRs

There is no existing issue that I saw that is requesting this functionality.

What does your PR implement? Be specific.

This PR uses matplotlib to implement a plotMA() function.
The MA plot shows the mean of the normalized counts versus the log2 fold changes for all genes tested. Genes that are significantly differentially expressed (based on some user-specified threshold) are colored red; the non-significant genes are colored grey. This plot can be helpful to look at LFC shrinkage effects.

This functionality tries to copy the version of plotMA() in R. Example:

A couple of considerations I wanted to flag:

1. Rows that return NaN in padj are colored grey and ultimately ignored (let me know if you'd like me to change this behavior, I think in R they are also ignored?)
2. I modify the results_df to add a new column, color. This may not be preferable, and I'd be grateful for input to clean this up.
3. There is some overlap between this and make_scatter() -- if it's helpful, I could potentially generalize make_scatter() to work with plot_dispersions() and MAplot()

[ghost]

hello @emmcauley , thanks a lot for your PR!

The tests are currently not passing due to the lack of an extended summary in the numpy doc style, meaning a longer sentence than the current description. Can you please add one?

Regarding the actual content, we are a bit short on staff at the moment, but we'll come back to you in a few weeks!

[emmcauley]

@mandreux-owkin, thanks for this context -- I've updated the extended summary and it looks like the checks are passing now. Let me know if you need anything else and/or want to see any changes before merging.

[BorisMuzellec]

Hi @emmcauley, thanks a lot for this new PR, and sorry for replying so late!

I made a couple of changes:

• I changed the code in utils.py to avoid adding a column to results_df with the colors,
• I added a wrapper method in DeseqStats,
• I added a pytest,
• and I did a bit of refactoring here and there.

Happy to merge if you're happy with the current state of this PR.

[emmcauley]

Sorry for the delay, thanks @BorisMuzellec, looks great! I think the use of p < thresh is more consistent with the R documentation (as opposed to p <= thresh).