Optimize some Mutect-related tools #5073
Conversation
| @@ -77,7 +77,7 @@ private void initializeTruthVariantsIfNecessary() { | |||
| } | |||
|
|
|||
| if (truthVariants == null) { | |||
| truthVariants = new FeatureDataSource<>(new FeatureInput<>(truthVariantsFile, "truth"), CACHE_LOOKAHEAD, VariantContext.class); | |||
| truthVariants = new FeatureDataSource<>(new FeatureInput<>(truthVariantsFile, "truth"), CACHE_LOOKAHEAD, VariantContext.class, cloudPrefetchBuffer, cloudIndexPrefetchBuffer); | |||
There was a problem hiding this comment.
@davidbenjamin You should make this same change to the evalVariants FeatureDataSource.
There was a problem hiding this comment.
Thanks @cmnbroad, did it.
davidbenjamin
force-pushed
the
db_m2_optimizations
branch
from
d6da091 to
518b9b7
Compare
Codecov Report
@@ Coverage Diff @@
## master #5073 +/- ##
===============================================
+ Coverage 86.385% 86.419% +0.034%
- Complexity 28822 29084 +262
===============================================
Files 1791 1791
Lines 133561 134336 +775
Branches 14902 15138 +236
===============================================
+ Hits 115377 116092 +715
- Misses 12791 12808 +17
- Partials 5393 5436 +43
|
| @@ -55,14 +53,15 @@ | |||
| * gatk GetPileupSummaries \ | |||
| * -I tumor.bam \ | |||
| * -V common_biallelic.vcf.gz \ | |||
| * -V common_biallelic.vcf.gz \ | |||
There was a problem hiding this comment.
Whoops -- one is -V and one should be -L. I have to admit this is kind of clumsy but it's not utterly redundant, at least. For example, one might want to use some custom interval list (-L) of common sites but grab the allele frequencies from gnomAD (-V).
Fixed it and expanded the javadoc a bit.
|
@LeeTL1220 back to you. |
There was a problem hiding this comment.
@davidbenjamin One more doc comment, then feel free to merge.
| @@ -65,6 +65,17 @@ | |||
| * -O pileups.table | |||
| * </pre> | |||
| * | |||
| * Although the sites (-L) and variants (-V) resources will often be identical, this need not be the case. For example, | |||
There was a problem hiding this comment.
Typically these files would only be different if you were trying to use a subset of the variants? I.e. -L was a subset of -V? If so, can you add that to the doc?
davidbenjamin
force-pushed
the
db_m2_optimizations
branch
from
5f82c7e to
351582c
Compare
@LeeTL1220 The first commit makes all concordance tools and MC3/M2 vcf merge much faster in Firecloud -- tasks that have been taking an hour will take a few minutes. The second commit makes
GetPileupSummaries, hence the contamination task much faster. Previously that tool has cached all reads for the 100,000 bases around each site, so basically it had to do a whole bam's worth of I/O just to get ~60,000 pileups.