97c5ed4feb
Merge branch 'master' of ssh://gsa4.broadinstitute.org/humgen/gsa-scr1/gsa-engineering/git/unstable
2012-08-06 20:22:31 -04:00
238d55cb61
Fixes for running HaplotypeCaller with reduced reads: a) minor refactoring, pulled out code to compute mean representative count to ReadUtils, b) Don't use min representative count over kmer when constructing de Bruijn graph - this creates many paths with multiplicity=1 and makes us lose a lot of SNP's at edge of capture targets. Use mean instead
2012-08-06 20:22:12 -04:00
d85b38e4da
Updating HaplotypeCaller integration tests
2012-08-06 12:02:19 -04:00
b8709d8c67
Merge branch 'master' of ssh://gsa2.broadinstitute.org/humgen/gsa-scr1/gsa-engineering/git/unstable
2012-08-06 11:41:28 -04:00
973d1d47ed
Merging together the computeDiploidHaplotypeLikelihoods functions in the HaplotypeCaller's LikelihoodEngine so they both benefit from the ReducedRead's RepresentativeCount
2012-08-06 11:40:07 -04:00
b7eec2fd0e
Bug fixes related to the changes in allele padding. If a haplotype started with an insertion it led to array index out of bounds. Haplotype allele insert function is now very simple because all alleles are treated the same way. HaplotypeUnitTest now uses a variant context instead of creating Allele objects directly.
2012-08-05 12:29:10 -04:00
d2e8eb7b23
Fixed 2 haplotype caller unit tests: a) new interface for addReadLikelihoods() including read counts, b) disable test that test basic DeBruijn graph assembly, not ready yet
2012-08-03 14:26:51 -04:00
c3b6e2b143
Merge branch 'master' of ssh://gsa2.broadinstitute.org/humgen/gsa-scr1/gsa-engineering/git/unstable
2012-08-03 13:14:43 -04:00
ff80f17721
Using PathComparatorTotalScore in the assembly graph traversal does a better job of capturing low frequency branches that are inside high frequnecy haplotypes.
2012-08-03 13:14:37 -04:00
6f8e7692d4
Merge branch 'master' of ssh://gsa4.broadinstitute.org/humgen/gsa-scr1/gsa-engineering/git/unstable
2012-08-03 12:24:37 -04:00
9e25b209e0
First pass of implementation of Reduced Reads with HaplotypeCaller. Main changes: a) Active region: scale PL's by representative count to determine whether region is active. b) Scale per-read, per-haplotype likelihoods by read representative counts. A read representative count is (temporarily) defined as the average representative count over all bases in read, TBD whether this is good enough to avoid biases in GL's. c) DeBruijn assembler inserts kmers N times in graph, where N is min representative count of read over kmer span - TBD again whether this is the best approach. d) Bug fixes in FragmentUtils: logic to merge fragments was wrong in cases where there is discrepancy of overlaps between unclipped/soft clipped bases. Didn't affect things before but RR makes prevalence of hard-clipped bases in CIGARs more prevalent so this was exposed. e) Cache read representative counts along with read likelihoods associated with a Haplotype. Code can/should be cleaned up and unified with PairHMMIndelErrorModelCode, as well as refactored to support arbitrary ploidy in HaplotypeCaller
2012-08-03 12:24:23 -04:00
3ece4c4993
Merged bug fix from Stable into Unstable
2012-08-02 11:41:36 -04:00
cb8bc18aeb
Fix for error in HaplotypeCaller. HC has a UG argument collection for the UG engine but some of those arguments aren't appropriate to set.
2012-08-02 11:41:06 -04:00
9ac72dbd4d
Merged bug fix from Stable into Unstable
2012-08-01 10:56:45 -04:00
01265f78e6
Add sanity check and possible bug fix for forum user: if haplotypes cannot be created from given alleles when genotyping indels (e.g. too close to contig boundary, etc.) in pool mode, empty allele list, signifying site can't be genotyped
2012-08-01 10:50:00 -04:00
4a23f3cd11
Simple cleanup of pool caller code - since usage is much more general than just calling pools, AF calculation models and GL calculation models are renamed from Pool -> GeneralPloidy. Also, don't have users specify special arguments for -glm and -pnrm. Instead, when running UG with sample ploidy != 2, the correct general ploidy modules are automatically detected and loaded. -glm now reverts to old [SNP|INDEL|BOTH] usage
2012-07-31 16:34:20 -04:00
e00ed8bc5e
Cleanup BQSR classes
...
-- Moved most of BQSR classes (which are used throughout the codebase) to utils.recalibration. It's better in my opinion to keep commonly used code in utils, and only specialized code in walkers. As code becomes embedded throughout GATK its should be refactored to live in utils
-- Removed unncessary imports of BQSR in VQSR v3
-- Now ready to refactor QualQuantizer and unit test into a subclass of RecalDatum, refactor unit tests into RecalDatum unit tests, and generalize into hierarchical recal datum that can be used in QualQuantizer and the analysis of adaptive context covariate
-- Update PluginManager to sort the plugins and interfaces. This allows us to have a deterministic order in which the plugin classes come back, which caused BQSR integration tests to temporarily change because I moved my classes around a bit.
2012-07-31 08:11:03 -04:00
e6b326c189
Merge branch 'master' of ssh://gsa4.broadinstitute.org/humgen/gsa-scr1/gsa-engineering/git/unstable
2012-07-30 21:32:19 -04:00
6c9d3ec155
Remerge after changes to allele construction code. More cleanups/fixes to artificial read pileup provider
2012-07-30 21:32:03 -04:00
3dabb90eb0
Updating example active region walker integration test.
2012-07-30 21:26:16 -04:00
c2b57ee444
updating HC integration tests after these changes.
2012-07-30 12:41:40 -04:00
13591b169f
Merge branch 'master' of ssh://gsa2.broadinstitute.org/humgen/gsa-scr1/gsa-engineering/git/unstable
2012-07-30 12:13:24 -04:00
48b9495460
Fixes to the likelihood based LD calculation for deciding when to combine consecutive events.
2012-07-30 12:12:56 -04:00
9002758ede
Merge branch 'master' of ssh://gsa2.broadinstitute.org/humgen/gsa-scr1/gsa-engineering/git/unstable
2012-07-30 12:10:00 -04:00
7a73042cd3
Bug fix for the case of merging two VCs when a deletion deletes the padding base for a consecutive indel. Added unit test to cover this case.
2012-07-30 12:09:23 -04:00
5743694196
Merged bug fix from Stable into Unstable
2012-07-30 11:35:28 -04:00
79195b97a3
Adding categories for the remaining uncategorized walkers
2012-07-30 11:35:08 -04:00
5b9a1af7fe
Intermediate fix for pool GL unit test: fix up artificial read pileup provider to give consistent data. b) Increase downsampling in pool integration tests with reference sample, and shorten MT tests so they don't last too long
2012-07-30 09:56:10 -04:00
99b15b2b3a
Final checkpoint: all tests pass. Note that there were bugs in the PoolGenotypeLikelihoodsUnitTest that needed fixing and eventually led to my needing to disable one of the tests (with a note for Guillermo to look into it). Also note that while I have moved over the GATK to use the new non-null representation of Alleles, I didn't remove all of the now-superfluous code throughout to do padding checking on merges; we'll need to do this on a subsequent push.
2012-07-29 01:07:59 -04:00
2b1b00ade5
All integration tests and VC/Allele unit tests are passing
2012-07-27 17:03:49 -04:00
beb7610195
Resolving merge conflicts
2012-07-27 15:52:02 -04:00
27e7e11ec0
Allele refactoring checkpoint #3 : all integration tests except for PoolCaller are passing now. Fixed a couple of bugs from old code that popped up during md5 difference review. Added VariantContextUtils.requiresPaddingBase() method for tools that create alleles to use for determining whether or not to add the ref padding base. One of the HaplotypeCaller tests wasn't passing because of RankSumTest differences, so I added a TODO for Ryan to look into this.
2012-07-27 15:48:40 -04:00
22bb4804f0
HaplotypeCaller now use an excessive number of high quality soft clips as a triggering signal in order to capture both end points of a large deletion in a single active region.
2012-07-27 12:44:02 -04:00
a0890126a8
ActiveRegionWalker's isActive function returns a results object now instead of just a double.
2012-07-27 11:01:39 -04:00
baf3e33730
Allele refactoring checkpoint 2: all code finally compiles, AD and STR annotations are fixed, and most of the UG integration tests pass.
2012-07-26 23:27:11 -04:00
35e803e110
Merged bug fix from Stable into Unstable
2012-07-26 14:00:04 -04:00
4f741b4cd7
Smoothing in the BQSR bins should be one error observation and one non-error observation.
2012-07-26 13:59:02 -04:00
2ae890155c
Improvements to indel calling in pool caller: a) Compute per-read likelihoods in reference sample to determine wheter a read is informative or not. b) Fixed bugs in unit tests. c) Fixed padding-related bugs when computing matches/mismatches in ErrorModel, d) Added a couple of more integration tests to increase test coverage, including testing odd ploidy
2012-07-26 13:43:00 -04:00
a694d1b5de
Merge branch 'master' into allelePadding
2012-07-26 01:53:14 -04:00
32516a2f60
Initial checkpoint commit of VariantContext/Allele refactoring. There were just too many problems associated with the different representation of alleles in VCF (padded) vs. VariantContext (unpadded). We are moving VC to use the VCF representation. No more reference base for indels in VC and no more trimming and padding of alleles. Even reverse trimming has been stopped (the theory being that writers of VCF now know what they are doing and often want the reverse padding if they put it there; this has been requested on GetSatisfaction). Code compiles but presumably pretty much all tests with indels with fail at this point.
2012-07-26 01:50:39 -04:00
7eb3f54750
Added category docs for the remaining public walkers (I think I got them all). I removed a couple of totally unnecessary walkers.
2012-07-25 21:40:28 -04:00
0a98a6aa8d
Adding extraDocs tag per Mauricio's request
2012-07-25 18:23:18 -04:00
05fa377a8e
Adding GATK categories to standard walkers. Will add to remaining walkers after the next successful release (so that I can see which walkers are public and still need it).
2012-07-25 16:05:47 -04:00
a5721a8846
Context covariate optimizations were not suited for multiple threads, so I removed them (since that ended up being much, much easier than trying to make the covariates thread local). Added -nt 2 layer to BQSR integration tests to confirm that it now works with multiple threads.
2012-07-25 13:38:07 -04:00
675ccab2fa
Renaming BQSR to BaseRecalibrator
2012-07-23 10:17:17 -04:00
2e486d83e2
Updating HaplotypeCaller docs and expanding integration tests.
2012-07-23 10:05:42 -04:00
df965d4a5a
Fixing BQSR integration test
2012-07-21 11:11:45 -04:00
116885a450
Removed the "Walker" suffix from all walkers that had it.
...
* Did not touch archived walkers... those can be named whatever.
* Kept abstract classes that end in Walker untouched (e.g. LocusWalker, ReadWalker, ...)
* Renamed a few inner classes due to conflict when stripping off Walker from their outer classes: ContigStats, FlagStats and FastaStats.
2012-07-20 17:27:11 -04:00
1592841c93
New function for merging nearby events into MNPs or complex substitutions. Added extensive unit tests.
2012-07-19 13:16:33 -04:00
c16f9f2f15
a) Use new method to check for GATK Like, b) minor improvements to indel pool caller (more to come): brain-dead, quick way to limit number of alt alleles to genotype. We can't process too many alt alleles because of the combinatorial explosion of GL values with high ploidy, and some STR validation targets had up to 12 alt alleles, resulting of GL vectors of > 1e8 elements. Can't use pileup elements since typically not many alleles will be in one pileup, and different alleles will appear in different samples, TBD a nicer solution. c) Commit to posterity scala script for large scale validation calling, still work in progress
2012-07-19 10:24:08 -04:00
5f5edeca63
Reverting move of BQSR tests to public, as per DR's email
2012-07-19 10:02:05 -04:00
9c1ab1b0c0
Move BQSR integration test and its dependent files into public; previously there was a protected->private dependency.
2012-07-18 21:11:33 -04:00
74e153ff4a
FisherStrand now uses RankSumTest isUsableBase to decide if a read should be included in testing
...
-- Previously used hardcoded MAPQ > 20 && QUAL > 20 but now uses isUsableBase
-- Updating MD5s as appropriate
2012-07-18 16:07:47 -04:00
e4db8dde91
Enabled a whole other bunch of integration tests for BQSRv2. While I was there I also changed the default context size for indels to 3 (from 8) since that's what works best in the current implementation (as suggested by Ryan). At this point, all of the new core tools (ReduceReads, BQSRv2, HaplotypeCaller, UG extensions) have been moved over to protected and should be stable. Looks like we are pretty much ready for GATK 2.0!
2012-07-17 23:36:43 -04:00
731bbba2e6
Bug fixes for integration test, use correct new UG syntax
2012-07-17 16:57:59 -04:00
40b8c7172c
Pool Caller refactoring in preparation of GATK 2.0: a) PoolCallerUnifiedArgumentCollection disappeared, and arguments moved to UnifiedArgumentCollection. b) PoolCallerWalker is no longer needed and redundant, all functionality subsumed by UG. UG now checks if GATK is lite - if so, don't allow ploidy > 2. c) Moved pool classes from private to protected. d) Changed the way to specify ploidy. Instead of specifying samples per pool and having ploidy = 2*samplesPerPool, have user specify ploidy directly, which is cleaner. Update tests accordingly. We can now call triploid seedless grape genotypes correctly in theory. e) Renamed argument -reference to -reference_sample_calls since the former is ambiguous and it's not clear what it refers to.
2012-07-17 15:27:04 -04:00
b0d99fd10d
Merge branch 'master' of ssh://gsa2.broadinstitute.org/humgen/gsa-scr1/gsa-engineering/git/unstable
2012-07-17 15:12:28 -04:00
305db8c0d1
Total rewrite of the isGATKLite() functionality with help of Khalid/David. PluginManager was not working for us.
2012-07-17 15:11:03 -04:00
bf2d5efe4d
Moving HaplotypeCaller integration and unit tests over to protected as well.
2012-07-17 14:51:26 -04:00
c55934043e
Moving HaplotypeCaller from private to protected
2012-07-17 14:41:19 -04:00
3a64398d07
Cleaned up the isGATKLite check
2012-07-17 12:46:16 -04:00
62c5228048
1) Revert previous change - indel recalibration is turned on by default and users of the Lite version will need to turn it off to avoid a User Error. 2) Implemented the engine.isGATKLite() method.
2012-07-17 12:23:40 -04:00
40618ac471
A bunch of BQSR changes: 1) by default we do not emit indel quals, but they can be turned on with --enable_indel_quals. 2) We check whether or not we are running in Lite mode (not done yet) and if so and the user is trying to recalibrate indels, we throw a User Error (not supported). 3) Like v1 we now allow the user to set the qual value below which we don't recalibrate (this was the remaining source of differences in the v1 vs. v2 plots).
2012-07-17 10:52:43 -04:00
d5b3a2eabf
Merge branch 'master' of ssh://gsa2.broadinstitute.org/humgen/gsa-scr1/gsa-engineering/git/unstable
2012-07-17 00:32:53 -04:00
f657b8bda8
Complete overhaul of the BQSRv2 integration tests. Much more comprehensive. Still need to deal with a few tests that need some modifications before I'm done, but I'll take care of that sometime tomorrow.
2012-07-17 00:32:34 -04:00
0a89adbcdb
Add utility decorators so that classes can tell you which package source they come from if they want to (suggested by Khalid). Using those decorators, we can easily pull out the BQSR updateDataForPileupElement() method into a standard RecalibrationEngine and an AdvancedRecalibrationEngine and use the protected one (AdvancedRE) if available (otherwise, the public one).
2012-07-16 15:34:50 -04:00
52baac1e16
Move BQSRv2 into public and v1 into the archive.
2012-07-16 14:23:38 -04:00
6c6a324583
Loosen a restriction on isOriginalRead()
...
* no longer needs to satisfy ReadAndIntervalOverlap.OVERLAP_CONTAINED
2012-07-16 14:07:10 -04:00
d7bf74fb7e
Updating default value for -mindel to the one used by Khalid in the pipeline and me in my tests.
2012-07-10 02:04:26 -04:00
6c17c50fa2
Updates to ReduceReads
...
* Added optional parameter to not hard clip on the interval border
* Made not clipping the default behavior (hence integration tests changed)
* Updated integration tests.
2012-07-09 13:46:51 -04:00
12d1c594df
Moving ReduceReads into protected
2012-06-25 17:01:33 -04:00
9c6bccfd8b
build system overhaul
...
* Added support for a protected directory whose contents are only made public in binary form
* Simplified and reorganized build.xml to improve readability and maintainability
* build.xml now autodetects most build properties:
-Includes private/protected if they exist
-No more STING_BUILD_TYPE or specialized targets for public-only, etc.
* Build targets have changed! There are now two main build options:
"ant" build everything (GATK and Queue)
"ant gatk" build just the GATK
It was too hard to build everything before -- now it is the default.
* To run tests with debugging, use -Dtest.debug=true -Dtest.debug.port=XXXX on the command line.
Much better than the old comment/uncomment method!
2012-05-17 15:16:29 -04:00
Guillermo del Angel
Ryan Poplin
Mark DePristo
Eric Banks
Mauricio Carneiro
Joel Thibault
David Roazen