Changing defaults in mitochondria mode for M2 filtering rehaul

[meganshand]

After #5688 we need to adjust some of the default values when in mitochondria mode. I tested these parameters with the mixture samples and it looked good. In particular note that the TLOD divided by depth filter is no longer needed, so the default would now be 0.

Again, would love to get this in before the release on Tuesday (fyi @droazen)

@ldgauthier @davidbenjamin @jsotobroad

[codecov-io]

Codecov Report

Merging #5827 into master will decrease coverage by <.01%.
The diff coverage is 50%.

@@             Coverage Diff              @@
##             master    #5827      +/-   ##
============================================
- Coverage     87.04%   87.03%   -0.01%     
+ Complexity    32153    32149       -4     
============================================
  Files          1975     1974       -1     
  Lines        147415   147413       -2     
  Branches      16225    16225              
============================================
- Hits         128315   128307       -8     
- Misses        13185    13188       +3     
- Partials       5915     5918       +3

Impacted Files	Coverage Δ	Complexity Δ
...e/hellbender/utils/variant/GATKVCFHeaderLines.java	94.91% <ø> (-0.03%)	11 <0> (ø)
...ute/hellbender/utils/variant/GATKVCFConstants.java	75% <ø> (ø)	4 <0> (ø)	⬇️
...lkers/mutect/filtering/Mutect2FilteringEngine.java	97.91% <ø> (-0.03%)	42 <0> (ø)
...r/tools/walkers/mutect/Mutect2IntegrationTest.java	87.56% <ø> (-0.03%)	89 <0> (ø)
.../mutect/filtering/M2FiltersArgumentCollection.java	93.1% <0%> (-6.9%)	6 <4> (+4)
...kers/mutect/clustering/SomaticClusteringModel.java	99.35% <100%> (ø)	65 <1> (ø)	⬇️
...utils/smithwaterman/SmithWatermanIntelAligner.java	50% <0%> (-30%)	1% <0%> (-2%)
...e/hellbender/engine/filters/ReadFilterLibrary.java	94.56% <0%> (-0.95%)	1% <0%> (ø)
...nder/engine/filters/ReadFilterLibraryUnitTest.java	100% <0%> (ø)	59% <0%> (+1%)	⬆️
...ithwaterman/SmithWatermanIntelAlignerUnitTest.java	60% <0%> (ø)	2% <0%> (ø)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fb2b5a2...6cc5267. Read the comment docs.

[davidbenjamin]

@meganshand This looks good. Do you want to delete the LodOddsOverDepthFilter and remove its threshold from M2FilteringArgumentCollection in this PR as well?

[meganshand]

I'll defer to @ldgauthier on that question: Should we completely delete the lod divided by depth filter now that the new filtering strategy seems to make it obsolete? I've only run with these changes on the mixtures, so I don't have a good sense if this is the kind of thing we'll need in the future or not?

[ldgauthier]

In the latest filtering paradigm, how would somebody who only wanted variants with really high quality bases change the default parameters?

[ldgauthier]

Once we resolve the filter retention this is good to go. It might be worth noting where the mito priors came from for posterity.

[ldgauthier]

These are close-ish to the autosomal priors -- where did they come from?

[meganshand]

Back of the envelope calculation. I'll add a comment.

[davidbenjamin]

In the latest filtering paradigm, how would somebody who only wanted variants with really high quality bases change the default parameters?

You could decrease f-score-beta (default 1.0) to bias the threshold optimization in favor of precision versus sensitivity.

[ldgauthier]

In the latest filtering paradigm, how would somebody who only wanted variants with really high quality bases change the default parameters?

You could decrease f-score-beta (default 1.0) to bias the threshold optimization in favor of precision versus sensitivity.

Okay. Rip it out.

[meganshand]

@ldgauthier or @davidbenjamin I would appreciate one more quick look now that the lod divided by depth filter has been deleted.

[davidbenjamin]

Still 👍 from me.