Normalize results by fraction of reads in some MultiQC plots

[uniqueg]

Is your feature request related to a problem? Please describe.
Some MultiQC plots (e.g., from FastQC, tin-score-calculation, STAR), report absolute numbers of reads rather than fractions of reads. While it is useful to know the sequencing depths of the different samples/libraries, it is perhaps not very informative/helpful to represent these differences in multiple plots, as this makes it hard to compare key metrics across samples, e.g., the fraction of (uniquely) mapped reads across samples. This is especially problematic when these plots are exported and put in publications as static images.

Describe the solution you'd like
Have one plot that reports sequencing depths/library sizes for each sample as bar plots. Have all other plots report fractions of reads instead.

Additional context
It may not be possible to configure MultiQC in a way that read fractions are visualized instead of read numbers. In that case, it is probably too much of a hassle to implement the desired solution.

[uniqueg]

I am realizing that some of these are already available (for STAR, FastQC). For Cutadapt it may be too much of a hassle to change the existing MultiQC plugin. However, for the tin-score-calculation plugin, this may still be a reasonable feature.

[AngryMaciek]

I would advise against adjusting official plugins.
Sure, we can modify tin however we like, but what do you exactly propose?
I can think of 2 things: get rid of the peak at 0, re-label y-axis to %.
These should be rather minor changes...
If that is on your mind could you please rename the issue and assign proper elements from the right bar? :)

[dominikburri]

My understanding of TIN scores is that it measures for each transcript its integrity.
Now, if you want to display the frequency instead of the histogram, I see possibly some problems arising. If a sample does not cover a lot of transcripts but with good quality it will not be distinguishable from a sample covering all transcripts with good quality. Right now the samples will be distinguishable, as the total count will be lower and visible as a line below all the other good quality samples. So, not only the mode of the distribution but also the height is relevant.

[AngryMaciek]

This is the repo with the plugins:
https://github.com/zavolanlab/multiqc-plugins

In order to adjust which TIN scores are plotted one needs to modify this script:
https://github.com/zavolanlab/multiqc-plugins/blob/dev/modules/tin_score/tin_score.py

It would be nice to get a decision here - do we just remove 0 from TIN scores? Are ALFA and zpca OK?

[mkatsanto]

Have TIN score as
-bargraph with or without zeros
-for the normalisation there would need to be an update on the tin-score repo and then resubmission to multiqc