(SV) Misc. improvements to the new location inference and type interpretation tool #4562

SHuang-Broad · 2018-03-22T20:26:01Z

Output bam instead of sam for assembly alignments
Instead of creating directory, new interpretation tool writes files (behavior consistent with current interpretation tool)
Prefix with sample name for output files' names
Add INSLEN annotation when there's INSSEQ
Clarify the boundary between AlignedContig and AssemblyContigWithFineTunedAlignments
Increase test coverage for AssemblyContigAlignmentsConfigPicker

Up to date plans for more cleanups and improvements posted in #4111

codecov-io · 2018-03-23T21:16:24Z

Codecov Report

Merging #4562 into master will decrease coverage by 18.083%.
The diff coverage is 42.373%.

@@               Coverage Diff                @@
##              master     #4562        +/-   ##
================================================
- Coverage     79.819%   61.737%   -18.083%     
+ Complexity     16999     13288      -3711     
================================================
  Files           1066      1062         -4     
  Lines          61876     61725       -151     
  Branches       10007      9993        -14     
================================================
- Hits           49389     38107     -11282     
- Misses          8578     20318     +11740     
+ Partials        3909      3300       -609

Impacted Files	Coverage Δ	Complexity Δ
...ender/tools/spark/sv/utils/GATKSVVCFConstants.java	`75% <ø> (ø)`	`1 <0> (ø)`	⬇️
.../discovery/alignment/ContigAlignmentsModifier.java	`85.535% <ø> (ø)`	`41 <0> (ø)`	⬇️
...ce/AssemblyContigAlignmentSignatureClassifier.java	`12.366% <0%> (-46.553%)`	`17 <0> (-13)`
.../DiscoverVariantsFromContigAlignmentsSAMSpark.java	`27.957% <0%> (-65.521%)`	`8 <0> (-14)`
...v/discovery/inference/BreakpointComplications.java	`60.311% <100%> (-0.118%)`	`20 <0> (ø)`
...der/tools/spark/sv/utils/GATKSVVCFHeaderLines.java	`82.54% <100%> (+0.282%)`	`7 <0> (ø)`	⬇️
.../sv/discovery/inference/CpxVariantInterpreter.java	`64.925% <100%> (-5.97%)`	`24 <0> (-1)`
...ls/spark/sv/discovery/alignment/AlignedContig.java	`92.453% <100%> (+2.798%)`	`22 <8> (+1)`	⬆️
...ry/inference/CpxVariantInducingAssemblyContig.java	`83.178% <100%> (+0.467%)`	`24 <0> (ø)`	⬇️
.../sv/StructuralVariationDiscoveryPipelineSpark.java	`32.394% <15%> (-56.412%)`	`0 <0> (-10)`
... and 455 more

cwhelan

Looks fine, just some minor comments. My biggest point is that it'd be nice to treat the input directory as a first class object and handle it programmatically.

cwhelan · 2018-03-26T19:18:23Z

.../org/broadinstitute/hellbender/tools/spark/sv/StructuralVariationDiscoveryPipelineSpark.java

@@ -110,14 +111,14 @@
            = new DiscoverVariantsFromContigsAlignmentsSparkArgumentCollection();
    @Argument(doc = "sam file for aligned contigs", fullName = "contig-sam-file")
    private String outputAssemblyAlignments;
-    @Argument(doc = "filename for output vcf", shortName = StandardArgumentDefinitions.OUTPUT_SHORT_NAME,
+    @Argument(doc = "prefix for output vcf; sample name will be appended after the provided argument",


Is this assuming that if you want files to appear in a directory then the argument should end with a trailing slash? I'd like to avoid that if possible. Can we handle this with eg the Path API?

I.e. Let's make this actually call for a directory. Then we should validate that the directory actually exists and create it if it's not there. Finally we can create the files in that directory using an API call, like Paths.resolve(outputDirectory, filename) for each filename we want.

is the commit 2fd080d similar to what you have in mind?

cwhelan · 2018-03-26T19:18:54Z

.../org/broadinstitute/hellbender/tools/spark/sv/StructuralVariationDiscoveryPipelineSpark.java

-            " the directory contains multiple VCF's for different types and record-generating SAM files of assembly contigs,",
-            fullName = "exp-variants-out-dir", optional = true)
-    private String expVariantsOutDir;
+    @Argument(doc = "prefix to output files of our prototyping breakpoint and type inference tool in addition to the master VCF;",


remove "in addition to the master VCF" from this doc line

cwhelan · 2018-03-26T19:19:11Z

.../org/broadinstitute/hellbender/tools/spark/sv/StructuralVariationDiscoveryPipelineSpark.java

-            fullName = "exp-variants-out-dir", optional = true)
-    private String expVariantsOutDir;
+    @Argument(doc = "prefix to output files of our prototyping breakpoint and type inference tool in addition to the master VCF;",
+            fullName = "exp-variants-out-prefix", optional = true)


let's keep this called directory according to my comment above

cwhelan · 2018-03-26T19:20:35Z

.../org/broadinstitute/hellbender/tools/spark/sv/StructuralVariationDiscoveryPipelineSpark.java

@@ -163,8 +164,9 @@ protected void runTool( final JavaSparkContext ctx ) {
        if(parsedAlignments.isEmpty()) return;

        final Broadcast<SVIntervalTree<VariantContext>> cnvCallsBroadcast = broadcastCNVCalls(ctx, headerForReads, discoverStageArgs.cnvCallsFile);
+        final String outputPrefixWithSampleName = outputPrefix + (outputPrefix.endsWith("/") ? "" : "_") + SVUtils.getSampleId(headerForReads);


if the user accidentally leaves off the trailing slash the files will not go into the directory requested but will be in its parent with a confusing name, i think.

right. thanks for catching that. corrected

cwhelan · 2018-03-26T19:23:25Z

...te/hellbender/tools/spark/sv/discovery/SvDiscoverFromLocalAssemblyContigAlignmentsSpark.java

-        SVFileUtils.writeSAMFile(outputDir+"/"+rawTypeString+".sam", splitLongReads.collect().iterator(),
-                headerBroadcast.getValue(), false);
+                .map(read -> read.convertToSAMRecord(header));
+        header.setSortOrder(SAMFileHeader.SortOrder.queryname);


Why sort these by queryname and not by position? I could be convinced either way but this requires an extra step to load the file into IGV.

that was because I basically used Ribbon for reviewing those alignments,
which is definitely easier with SAM compared to with BAM.
Now changed to adding an parameter to the utility

if BAM, sort by coordinate order.

if SAM, sort by query name.

cwhelan · 2018-03-26T19:31:21Z

.../org/broadinstitute/hellbender/tools/spark/sv/StructuralVariationDiscoveryPipelineSpark.java

@@ -391,7 +391,7 @@ private void experimentalInterpretation(final JavaSparkContext ctx,
                        final String contigName = AlignedAssemblyOrExcuse.formatContigName(alignedAssembly.getAssemblyId(), contigIdx);
                        final List<AlignmentInterval> arOfAContig
                                = getAlignmentsForOneContig(contigName, contigSequence, allAlignments.get(contigIdx), refNames, header);
-                        return new AlignedContig(contigName, contigSequence, arOfAContig, false);
+                        return new AlignedContig(contigName, contigSequence, arOfAContig);


Can we change this variable name? arOfAContig is not obvious any more. Maybe alignmentsForContig?

cwhelan · 2018-03-26T19:35:09Z

...tute/hellbender/tools/spark/sv/discovery/alignment/AssemblyContigAlignmentsConfigPicker.java

-     * or less summed mismatches if still tie
-     * first
+     * implement ordering that
+     *      prefer the configuration with less alignments, then


cwhelan · 2018-03-26T19:36:27Z

...tute/hellbender/tools/spark/sv/discovery/alignment/AssemblyContigAlignmentsConfigPicker.java

-     * first
+     * implement ordering that
+     *      prefer the configuration with less alignments, then
+     *      prefer the configuration less summed mismatches if still tie


"and prefers the configuration with a lower number of summed mismatches in case of a tie"

cwhelan · 2018-03-26T19:40:44Z

src/test/java/org/broadinstitute/hellbender/utils/StringSplitSpeedUnitTest.java

@@ -19,7 +19,7 @@
            "Italiam, fato profugus, Laviniaque venit " +
            "litora, multum ille et terris iactatus et alto " +
            "vi superum saevae memorem Iunonis ob iram; " +
-            "multa quoque et bello passus, dum conderet urbem, " +
+            "multa quoque et bello passus, testConfigurationSorting conderet urbem, " +


was this maybe a global search/replace change? My high school latin is pretty rusty but I don't think testConfigurationSorting is a word that can replace a conjunction like 'dum'. :)

definitely better than mine, who initially named a function "dum" 😰

bug-fix and improvement commit BUG-FIX: * fix a bug in ContigAlignmentsModifer.splitGappedAlignment() that was keeping the aligner score for the split child alignments while using NO_NM, it now uses NO_AS IMPROVEMENT 1: * new INSLEN annotation to accompany INSSEQ * assembly alignments are now output as BAM instead of SAM * sample name included in output VCF files * CLI now asks user to provide directory for VCF output and a boolean for optionally running experimental interpretation tool * experimental interpretation only writes BAM for ambiguous, incomplete, and misasseblysuspects IMPROVEMENT 2: clarify the boundary between the two classes AlignedContig and AssemblyContigWithFineTunedAlignments * AlignedContig now represents assembly contig, possibly unmapped, whose alignments are given by aligner as-is * AssemblyContigWithFineTunedAlignments represents contig whose alignments underwent selection and gap-split IMPROVEMENT 3: increase test coverage for AssemblyContigAlignmentsConfigPicker * add test for speedUpWhenTooManyMappings() * add test for sortConfigurations() * add test for splitGaps() and gappedAlignmentOffersBetterCoverage()

droazen · 2018-03-28T15:51:54Z

@SHuang-Broad The testSAMWriter_chr20.bam bam that you added here should probably have gone into git lfs (src/test/resources/large), rather than directly into the repo. It's too late to fix (once it's in master, it's in the git history forever), but in the future please try to store test data bigger than ~1-2 MB or so in lfs.

bug-fix and improvement commit BUG-FIX: * fix a bug in ContigAlignmentsModifer.splitGappedAlignment() that was keeping the aligner score for the split child alignments while using NO_NM, it now uses NO_AS IMPROVEMENT 1: * new INSLEN annotation to accompany INSSEQ * assembly alignments are now output as BAM instead of SAM * sample name included in output VCF files * CLI now asks user to provide directory for VCF output and a boolean for optionally running experimental interpretation tool * experimental interpretation only writes BAM for ambiguous, incomplete, and misasseblysuspects IMPROVEMENT 2: clarify the boundary between the two classes AlignedContig and AssemblyContigWithFineTunedAlignments * AlignedContig now represents assembly contig, possibly unmapped, whose alignments are given by aligner as-is * AssemblyContigWithFineTunedAlignments represents contig whose alignments underwent selection and gap-split IMPROVEMENT 3: increase test coverage for AssemblyContigAlignmentsConfigPicker * add test for speedUpWhenTooManyMappings() * add test for sortConfigurations() * add test for splitGaps() and gappedAlignmentOffersBetterCoverage()

SHuang-Broad added enhancement tests SV labels Mar 22, 2018

SHuang-Broad force-pushed the sh-misc-improvements branch 2 times, most recently from 8998c91 to 7e37a74 Compare March 23, 2018 19:55

cwhelan approved these changes Mar 26, 2018

View reviewed changes

SHuang-Broad force-pushed the sh-misc-improvements branch from a6cb572 to c202eb3 Compare March 27, 2018 21:05

SHuang-Broad merged commit 30669c3 into master Mar 27, 2018

SHuang-Broad deleted the sh-misc-improvements branch March 27, 2018 22:07

SHuang-Broad mentioned this pull request Mar 27, 2018

annotation "INSLEN" missing from SV VCF (and header) #4382

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

(SV) Misc. improvements to the new location inference and type interpretation tool #4562

(SV) Misc. improvements to the new location inference and type interpretation tool #4562

SHuang-Broad commented Mar 22, 2018

codecov-io commented Mar 23, 2018 •

edited

Loading

cwhelan left a comment

cwhelan Mar 26, 2018

SHuang-Broad Mar 27, 2018

cwhelan Mar 26, 2018

SHuang-Broad Mar 27, 2018

cwhelan Mar 26, 2018

cwhelan Mar 26, 2018

SHuang-Broad Mar 27, 2018

cwhelan Mar 26, 2018

SHuang-Broad Mar 27, 2018

cwhelan Mar 26, 2018

SHuang-Broad Mar 27, 2018

cwhelan Mar 26, 2018

SHuang-Broad Mar 27, 2018

cwhelan Mar 26, 2018

SHuang-Broad Mar 27, 2018

cwhelan Mar 26, 2018

SHuang-Broad Mar 27, 2018

droazen commented Mar 28, 2018

(SV) Misc. improvements to the new location inference and type interpretation tool #4562

(SV) Misc. improvements to the new location inference and type interpretation tool #4562

Conversation

SHuang-Broad commented Mar 22, 2018

codecov-io commented Mar 23, 2018 • edited Loading

Codecov Report

cwhelan left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

droazen commented Mar 28, 2018

codecov-io commented Mar 23, 2018 •

edited

Loading