Integrate PCA plot in multiqc report

The rules (pca_salmon, pca_kallisto) generate the PCA results for genes and transcripts. The plots should be integrated into multiqc and not as custom plots as we have them now. This involves:

• Add code to generate the plots in https://github.com/zavolanlab/multiqc-plugins
• Use the new Dockerfile in zarp and adjust Snakefile (wiring of rules)

changed milestone to %downstream

added To Do downstream labels

assigned to @bakma

added Doing label and removed To Do label

We are working on it, the issue is here:
https://github.com/zavolanlab/multiqc-plugins/issues/18

marked the checklist item Add code to generate the plots in https://github.com/zavolanlab/multiqc-plugins as completed

The plugin is ready; I merged a new minor version to master and released a tagged release; container was automatically pushed to DockerHub.

@gypas

I run zarp on the GCN4 test dataset with the new multiqc plugins image. In principle everything works fine but the report is not as we would like it to be. Please look at the attached archive: multiqc_summary.zip. The thing is that each interactive plot is by definition one subsection of the report and each plugin is a separate section. Thus, right now the zpca section is a clogged by all the plots altogether. We previously encountered a similar issue for the ALFA plots (take a look there) and solved it by encoding the unique information into the subsection name. It worked there since every plot was indeed a separate entity. Here, the problem is that the plots go in triplets: 1x2,2x3,3x1. You have any design ideas?

I agree that it does not look very good at the moment. One thing is to move the zpca plots at the end. This can happen by adjusting the following script: https://git.scicore.unibas.ch/zavolan_group/pipelines/zarp/-/blob/dev/workflow/scripts/zarp_multiqc_config.py

The second is that we need to add custom labels to show that the analysis is based on kallisto/salmon and at the gene/transcript level. I think we should check the options here: https://multiqc.info/docs/#customising-reports

@gypas

1. OK, I forgot about that configuration. Done!

2. Well... Yes... I agree... My point was: Under the report framework circumstances - do you have any desired outcome of our modifications in mind? The only thing I can imagine is to do the same as for ALFA module: parse the results subdirectory name and append it to the plot name (and subsection name, in consequence). under zpca in the results there are 4 subdirs: pca_kallisto_genes pca_kallisto_transcripts pca_salmon_genes pca_salmon_transcripts. This, however, would result in a long name, like: pca_kallisto_genes: PCA components: 1 & 2. But I see no other way around, to be honest.

If there is no other option then this is what we should do. Maybe if we add a new line after the first : it will look fine.

Something like this:

pca_kallisto_genes: 
PCA components: 1 & 2

@gypas

I must have had a blackout or sth... There is no need for any modifications whatsoever! We will just create 4 separate sections and point to these directories separately from the config (similarly as for cutadapt). Take a look here: multiqc_summary.zip

marked the checklist item Use the new Dockerfile in zarp and adjust Snakefile (wiring of rules) as completed

Ok, great. Looks good to me.

mentioned in merge request !82 (merged)

@bakma I added the requested tests. The new version is the following: https://github.com/zavolanlab/zpca/releases/tag/v0.8.3-1 Please let me know if there is any issue with that.

Thanks a lot @gypas, I have created an issue in the multiqc plugins repo:
https://github.com/zavolanlab/multiqc-plugins/issues/22

I hope Krish will be able to correct it soon. Then I push the image and Merge Request here again.

mentioned in commit 6f89145d

mentioned in commit 36c02328

Done in 6f89145d