56433ebf6b
Switched from LSF command line wrappers to JNA wrappers around the C API. Side effects:
...
- bsub command line is no longer fully printed out.
- extraBsubArgs hack is now a callback function updateJobRun.
Updated FullCallingPipelineTest to reflect latest changes to fullCallingPipeline.q.
Added a pipeline that tests the UGv2 runtimes at different bam counts and memory limits.
Updated VE packages that live in oneoffs to compile to oneoffs.
Added a hack to replace the deprecated symbol environ in Mac OS X 10.5+ which is needed by LSF7 on Mac.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4816 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-10 04:36:06 +00:00
d4d3170436
Support for '-L unmapped' in read walkers. DO NOT USE THIS PATCH YET. It has been
...
subjected to and passes cursory testing on one dataset (and all integration tests pass).
However, there's a small library of validation checks, and unit and integration tests
that must be added.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4813 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-09 19:51:48 +00:00
a2d6cef181
Weird corner condition fix in indel genotyper: if there are 2 consecutive locations on candidate sites to genotype, we can get both when calling getVariantContexts and if we are triggering on an extended event - this leads to confusion and we can end up picking the wrong one. So, we require start of the vc to be the same as the start of the ref locus to be sure.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4812 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-09 19:34:23 +00:00
722819688a
Minor utility improvements to ValidateBAQ
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4809 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-09 02:19:32 +00:00
a63bbb2fec
Optimized BAQ implementation. No longer does excessive amounts of copying of arrays. At this point I'm not 100% certain where additional performance improvements would come from
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4808 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-08 21:26:30 +00:00
db55b2b0c6
Better testing of BAQ. Now really handles soft clipped reads properly by doing an expensive copy operation :-( will need to be transformed to a ByteBuffer in the near future.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4807 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-08 17:37:00 +00:00
f1f01610f8
Remove the extra trailing tab at the end of the VCF ## header line. Unfortunately, this meant updating every freaking integration test.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4806 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-08 17:22:29 +00:00
16e1bbd380
Hidden command line option to control BAQ gap open penalty for testing by me and eric. ValidateBAQWalker has misc. useful improvements. PrintReads now adds BAQ tags on output, if requested.
...
BAQ has generally useful improvements. Refactor code to make it easier for BAQUnitTest to run. minBaseQuality enforced on output, as well as input now. Added BAQUnitTest that checks that the BAQ calculation is performing as expected. Still needs to be expanded significantly.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4804 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-08 01:01:39 +00:00
1b6bec8e6b
Trivial changes
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4803 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-07 20:06:54 +00:00
ca7810f11d
First major update of indel genotyper:
...
a) Really fix this time strand bias computation for indels, previous version was a partial fix only.
b) Change way in which we deal with bad bases at the edge of reads. Even if a base is soft clipped in CIGAR string, there may still be dangling bases with Q=2 that may throw off QUAL computation in some sites. So, we're stricter and we also trim off those bases off read edges even if they are not soft-clipped officially.
c) First feeble-minded attempt at runtime optimization - don't compute log and 10^base_qual every time. Rather, cache 10^-k/10 and log(1-10^-k/10) for all k <=60. This speeds up code about 4x.
d) Further optimization: don't compute log(10^x+10^y) but rather use softMax function recently put into ExactAFCalculationModel.
e) Skip bad reads where all Q=2 (sic)
f) Avoid log to lin and back to log conversions of genotype likelihoods - this was legacy code from back when exact model did stuff in linear domain. This improves precision overall.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4802 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-07 18:35:22 +00:00
e2d45ec2af
Make Indel Realigner exceptions related to not enough space on disk or a too low file-handle limit UserExceptions.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4801 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-07 16:37:31 +00:00
70980b659a
CombineVariants no longer requires rod_priority_string
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4800 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-07 15:39:43 +00:00
bc885b7bd0
Don't print debugging output.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4799 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-06 20:57:11 +00:00
c91712bd59
BAQ calculation refactoring in the GATK. Single -baq argument can be NONE, CALCULATE_AS_NECESSARY, and RECALCULATE. Walkers can control bia the @BAQMode annotation how the BAQ calculation is applied. Can either be as a tag, by overwriting the qualities scores, or by only returning the baq-capped qualities scores. Additionally, walkers can be set up to have the BAQ applied to the incoming reads (ON_INPUT, the default), to output reads (ON_OUTPUT), or HANDLED_BY_WALKER, which means that calling into the BAQ system is the responsibility of the individual walker.
...
SAMFileWriterStub now supports BAQ writing as an internal feature. Several walkers have the @BAQMode applied to this, with parameters that I think are reasonable. Please look if you own these walkers, though
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4798 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-06 20:55:52 +00:00
5d2c2bd280
Just refactoring into utils/baq directory
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4795 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-06 17:43:43 +00:00
80f32712dc
Tiny bug fix
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4793 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-05 18:48:33 +00:00
44feb4a362
Improved BAQ implementation. Now supports adding BAQ tags to reads on the fly with ADD_TAG_ONLY option. Caching fasta reader implementation, and changes throughout the system to enable this. Many performance improvements throughout the system due to better reference access patterns.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4792 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-05 18:29:39 +00:00
8901e63879
Cheap optimization: don't keep calculating the log of a constant. (How did I not catch this before?)
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4791 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-05 04:36:21 +00:00
bef48e7a42
For Chris, to make his life easier: iterate over all VCF records passed in looking for one with an ALT allele defined instead of assuming all records have one.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4789 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-05 02:23:38 +00:00
97c94176c0
Immediate, obvious bug fix to avoid blowing up on unmapped reads
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4788 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-04 20:43:39 +00:00
a5b3aac864
Engine-level BAQ calculation now available in the GATK [totally experimental right now]. -baq argument to disable (NONE), to only use the tags in the BAM (USE_TAG_ONLY), use the tag when present but calculate on the fly as necessary (CALCULATE_AS_NECESSARY), and to always recalculate (RECALCULATE_ALWAYS). BAQ.java contains the complete implementation, for those interested. ValidateBAQWalker is a useful QC tool for verifying the BAQ is correct. BAQSamIterator applies BAQ to reads, as needed, in the engine. Let me know if you encounter any problems. Before prime-time, needs a caching implementation of IndexedFastaReader to avoid loading *lots* of reference data all of the time
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4787 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-04 20:23:06 +00:00
b12cec4302
Added emitOnlyMNPs flag
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4785 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-03 20:34:17 +00:00
6d4ec7f9e7
Remove RefSeq INFO from MNPs since annotating them properly
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4784 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-03 19:03:35 +00:00
4719bbc772
Changed dontRequireSomeSampleHasDoubleAltAllele parameter to mean that merging should only start at a polymorphic site
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4783 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-03 17:52:56 +00:00
ec174dc0ba
As per Menachem's last commit, there's a minimally more efficient way of doing the MQ cap.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4782 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-03 16:37:08 +00:00
92cf7744a6
Set minMQ = max(minMQ, minBQ) for phasing since anyway we cap BQ by MQ; also, lowered MIN_BASE_QUALITY_SCORE for phasing to 17 (was previously 20)
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4781 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-03 16:31:13 +00:00
237ab1d489
1. As discussed in group meeting today, because we cap BQ by MQ, if MQ < minBQ then we filter the read.
...
2. Update to UGCalcLikelihoods for Chris: require a vcf bound to 'allele' to be provided so that we know exactly which alternate allele we should be calculating GLs for at each site. The user is warned when the VC is not biallelic or there are multiple records at a site.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4780 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-03 05:57:06 +00:00
da6a07ad3b
First round of critical fixes to indel genotyper (more to come tomorrow):
...
a) Avoid complete crash of caller that broke due to a recent refactoring by someone who must not be named <cough>EB<cough>... an integration test to avoid this in the future coming soon.
b) Fixed up strand bias computation for indels
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4779 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-03 02:48:09 +00:00
e09d6ee56b
write non-MNP VariantContexts records only once (where they start)
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4777 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-02 22:14:26 +00:00
1515bf6de9
Merged common VCF writing logic into phasing/WriteVCF.java
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4776 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-02 22:03:02 +00:00
4e62de4213
Added method getOriginalReadGroupId(): takes merged (in case of collision) read group id as reported by a read coming from the merged stream and returns this read's read group id as it was listed in the original input bam file.
...
IndelRealigner now uses this functionality to correctly un-mangle read group id's in --nWayOut mode (i.e. when we need to write reads into separate output bams with headers matching the original inputs).
Some hidden changes to IndelRealigner: purely testing and development, transparent to the users (hidden option added)
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4775 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-02 21:41:52 +00:00
e5282742f9
Bug fix in CountCovariates, skip over indel records as well as SNPs in the dbsnp file. CountCovariates is now called CountCovariatesWalker. I've always hated that the name was swapped.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4774 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-02 18:43:24 +00:00
0adf505b53
We no longer look at by-hapmap validation status in the VQSR because using the HapMap VCF file is higher quality. As a side effect we now support the dbsnp 132 vcf file. ApplyVariantCuts now requires that the input VCF rod bindings begin with input, matching the other VQSR walkers. Wiki updated with information about how to obtain the hapmap and 1kg truth sets.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4772 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-02 15:38:45 +00:00
99b942b0b4
Removing duplicated header args
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4770 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-01 20:16:53 +00:00
9ac0f98d0d
Fixed bug in retaining proper RefSeq records
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4768 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-01 18:39:02 +00:00
7caf666f48
For Sendu: add a hidden option to allow bams to come out unsorted. We've agreed to let him deal with sorting these puppies on his own.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4767 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-01 17:56:13 +00:00
3afa841a6a
Fixing docs
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4766 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-01 17:36:47 +00:00
6a6cdc1925
Adding minor usage docs
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4765 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-01 17:34:33 +00:00
0d1c905df3
Adding UGCalcLikelihoods and UGCallVariants so that GSA members can break up the calling process into separate steps (calculate the GLs and then call off of those) - useful for Chris's new batch merger. As the docs say, these are absolutely not supported or recommended for public use.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4764 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-01 17:32:26 +00:00
b4ef716aaf
As per Eric and Mark's suggestions, separated the segregating MNP merger (MergeMNPs) from the more general merger employed for annotation purposes (MergeSegregatingAlternateAlleles). Both use the same core MergePhasedSegregatingAlternateAllelesVCFWriter
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4763 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-01 16:42:08 +00:00
0892daddb0
Improvement for the TGEN folks: when running in the solid recal mode of SET_Q_ZERO_BASE_N, update the NM tag if one was present in the read to reflect the new N's in the read.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4761 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-01 04:36:44 +00:00
a22b1b04e6
SW-turbo. Kind of. This implementation is presumably equivalent to the old one (mathematically), but runs ~10 times faster: inner loops eliminated completely. The author of the original implementation should be sentenced to the galleys. Oh, that would be me...
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4760 348d0f76-0448-11de-a6fe-93d51630548a
2010-12-01 00:08:47 +00:00
2ac938fe4e
1)
...
Minor fixes to avoid crashes vs CG indel files:
- Add count for complex events, not just insertions and deletions
- Handle correctly cases of large indels falling out of bounds of histogram array: added a count of indels ouf of bounds and avoid exceptions.
2) Cosmetic fix for R script assessing UG calling performance: draw red y=x line on top of Simulated vs Estimated AC to get a better view of under/over-estimation of AC.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4758 348d0f76-0448-11de-a6fe-93d51630548a
2010-11-30 21:08:25 +00:00
af84462f3e
The dev team has decided to change the filter that is added to records that are set to monomorphic by Beagle. It no longer lists the reference allele. Added those filters to the header of the output VCF file. Finally, we no longer use R2=NaN values coming from Beagle.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4757 348d0f76-0448-11de-a6fe-93d51630548a
2010-11-30 17:19:54 +00:00
21256909bb
Not supported. I'm checking this in for Ryan only.
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4756 348d0f76-0448-11de-a6fe-93d51630548a
2010-11-30 16:59:18 +00:00
e21a66d876
Updated the Queue GATK generator and packaging to include more dependencies for fullCallingPipeline.q.
...
Set the -bigMemQueue in the FullCallingPipelineTest to GSA to avoid waiting for the week queue when it is busy.
Fixed the package definition of PipelineTest so that scalac won't recompile it every time.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4755 348d0f76-0448-11de-a6fe-93d51630548a
2010-11-30 15:29:40 +00:00
7f2ded0706
belated special case fix for Menachem; if the results of a BTI and BTIMR produce an empty interval list, exception out. This would be solved long term with better handling or empty and / or null interval lists. I'll add a JIRA
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4754 348d0f76-0448-11de-a6fe-93d51630548a
2010-11-30 05:49:20 +00:00
a181680814
We no longer require dbSNP files to be of the dbsnp rod-type; VCFs will do (provided they are bound to the name 'dbsnp')
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4753 348d0f76-0448-11de-a6fe-93d51630548a
2010-11-30 03:25:18 +00:00
8ffea42b75
about 10% improvement in SW alignment (and hence IndelRealigner!) speed by using c-style linearized array representation for matrices instead of java 2D arrays...
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4751 348d0f76-0448-11de-a6fe-93d51630548a
2010-11-30 00:06:50 +00:00
b03ac61e9d
consolidating the checking of the RMD sequence dictionary against the reference into a single function, and adding an integration test to test that empty VCFs pass (both the indexing and the seq dictionary validation).
...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4750 348d0f76-0448-11de-a6fe-93d51630548a
2010-11-30 00:01:56 +00:00
kshakir
hanna
delangel
depristo
ebanks
fromer
asivache
rpoplin
aaron