Dragen: v1.14 no longer looks for .bed files for coverage metrics

[iatol1]

Description of bug

Somewhere between v1.13 and v1.14, search_patterns.yaml was changed as follows:

https://github.com/ewels/MultiQC/blob/c247d077801bc1cd8d6e7e1f9078db5036bae6e1/multiqc/utils/search_patterns.yaml#L262-L267

vs.

https://github.com/ewels/MultiQC/blob/4702c8086ce95521d1e2c46d1a85ebfb41812d8e/multiqc/utils/search_patterns.yaml#L284-L291

This change means that MultiQC no longer looks for "*.target_bed_*.csv when plotting coverage metrics. It seems like a fix would be similar to what was done in #1290, but I am not sure.

The following regexes also seem relevant, but I don't know how they are used:

https://github.com/ewels/MultiQC/blob/4702c8086ce95521d1e2c46d1a85ebfb41812d8e/multiqc/modules/dragen/coverage_metrics.py#L292

https://github.com/ewels/MultiQC/blob/4702c8086ce95521d1e2c46d1a85ebfb41812d8e/multiqc/modules/dragen/coverage_hist.py#L144

https://github.com/ewels/MultiQC/blob/4702c8086ce95521d1e2c46d1a85ebfb41812d8e/multiqc/modules/dragen/coverage_per_contig.py#L166

If someone can give advice, I am happy to write a PR for a fix.

[ewels]

Could @vladsavelyev or @alexiswl or anyone else involved in #1829 / the regent DRAGEN updates please take a look at this?

I guess it's just that merging this old PR basically overwrote what was added by @vladsavelyev in #1363.

It would be good if someone with closer knowledge of this module could comment though please 🙏🏻

[alexiswl]

Thanks @iatol1 for picking this up, my guess is that this mammoth commit 08c5272 by @andrei-seleznev back in July 2021 to merge upstream master into the bnbowman master chose between the target and wgs coverage options when resolving conflicts.

There was a lot of code to digest through so pretty easy to miss. Since that merge commit would have both the bnbowman master and upstream master as parents, we wouldn't have seen this as a conflict when #1232 was finally merged.

I have access to a dragen instance so can try a fix on our side.

[alexiswl]

@iatol1 which version of dragen are you using as well? I will likely run these tests on dragen 4.0.3

[iatol1]

@alexiswl Yes, we are using Dragen 4.0.3 as well

[ckandoth]

The Dragen pipeline always generates *.wgs_*.csv files, and I don't see an option to disable it. They are not relevant for targeted sequencing, so it is preferred that MultiQC use *.target_bed_*.csv files when available and ignore the WGS ones.

The *.target_bed_*.csv files are only generated when a --vc-target-bed is specified or what we do is --qc-coverage-tag-1 target_bed --qc-coverage-region-1 exons.bed --qc-coverage-reports-1 cov_report. This latter method has the benefit of generating coverage metrics in targeted regions, but not limiting variants calls to the same regions.

[alexiswl]

The Dragen pipeline always generates .wgs_.csv files, and I don't see an option to disable it. They are not relevant for targeted sequencing, so it is preferred that MultiQC use .target_bed_.csv files when available and ignore the WGS ones.

@ckandoth thank you for clarifying

This latter method has the benefit of generating coverage metrics in targeted regions, but not limiting variants calls to the same regions.

That's a really good idea, thank you for sharing!

[ewels]

Hi all! Any progress on this front? I'm keen to release a patch release ASAP to fix a few bugs that have appeared, it would be great to be able to include this if possible..

[alexiswl]

Hi @ewels apologies, will get you something for this by the end of this week!

[alexiswl]

@iatol1 I have made a PR at #1854 (with some examples present too)

Please feel free to pass through any dragen csv files to test this with, otherwise you can checkout the branch at https://github.com/alexiswl/MultiQC/tree/bugfix/re-add-target-metrics-to-dragen-module.

Note that this still uses the wgs metrics, just also adds in the qc coverage metrics and target bed metrics tables.