Restore base quality filter code that got reverted in #4895

[takutosato]

This unintended change crept in with #4895.

@meganshand @davidbenjamin

[takutosato]

@meganshand please review!

[codecov-io]

Codecov Report

Merging #5123 into master will decrease coverage by 0.068%.
The diff coverage is 95%.

@@               Coverage Diff               @@
##              master     #5123       +/-   ##
===============================================
- Coverage     86.735%   86.668%   -0.068%     
+ Complexity     29312     29049      -263     
===============================================
  Files           1810      1809        -1     
  Lines         135549    134690      -859     
  Branches       15031     14934       -97     
===============================================
- Hits          117569    116733      -836     
+ Misses         12566     12545       -21     
+ Partials        5414      5412        -2

Impacted Files	Coverage Δ	Complexity Δ
...dorientation/CollectF1R2CountsIntegrationTest.java	100% <100%> (ø)	13 <1> (-3)	⬇️
...r/tools/walkers/mutect/Mutect2IntegrationTest.java	92.515% <100%> (+0.658%)	61 <1> (+1)	⬆️
...r/tools/walkers/mutect/Mutect2FilteringEngine.java	84.375% <66.667%> (+0.164%)	50 <0> (ø)	⬇️
...ellbender/tools/walkers/mutect/M2TestingUtils.java	91.304% <91.304%> (ø)	6 <6> (?)
...iscoverFromLocalAssemblyContigAlignmentsSpark.java	72.193% <0%> (-18.334%)	9% <0%> (-15%)
...ery/inference/SimpleNovelAdjacencyInterpreter.java	80.882% <0%> (-8.824%)	11% <0%> (-1%)
...institute/hellbender/utils/NucleotideUnitTest.java	90.698% <0%> (-4.795%)	34% <0%> (-91%)
...ls/ExtractOriginalAlignmentRecordsByNameSpark.java	86.364% <0%> (-4.545%)	9% <0%> (-1%)
...nalAlignmentRecordsByNameSparkIntegrationTest.java	86.667% <0%> (-3.958%)	7% <0%> (-2%)
...ute/hellbender/tools/walkers/vqsr/TrainingSet.java	73.333% <0%> (-3.333%)	4% <0%> (-1%)
... and 66 more

[takutosato]

@meganshand I added a very specific test to check for this bug that I introduced. I confirmed that the test fails before this change and passes after.

[meganshand]

Thanks for the test!

[meganshand]

Did you mean to make these changes? They look unrelated to the other test.

[sbamin]

HI @takutosato,

gatk/src/main/java/org/broadinstitute/hellbender/tools/walkers/mutect/Mutect2FilteringEngine.java

Lines 115 to 117 in 9df7b59

	if (baseQualityByAllele != null && baseQualityByAllele[indexOfMaxTumorLod + 1] < MTFAC.minMedianBaseQuality) {
	filterResult.addFilter(GATKVCFConstants.MEDIAN_BASE_QUALITY_FILTER_NAME);
	}

Looks like this could have impacted one or more of mutect2 somatic filtering (doc-11136). I believe bug was related to an argument, --min-median-base-quality, i.e., it was either ignoring this argument or (hope) taking a default value of 20 in Mutect2 and FilterMutectCalls tools.

• Is that the correct interpretation?
• Would this mandate running mutect2 from scratch or only FilterMutectCalls and subsequent filtration steps?

I am looking through our mutect2 filtered calls and filtering rules applied. Among PASS calls, so far, I don't see false positive with respect to FORMAT/MBQ < 20. Among filtered out calls, I don't have calls failing due to single filter but I need to look more into raw calls. Anyways, it would be of help to get your comments on possible impact of this bug in gatk version < 4.0.9.0 (I used 4.0.8.1).