c3c66e853cImprovements for Jason
depristo
2010-07-09 20:18:37 +0000
405be230d0Various code improvements based on FindBugs
ebanks
2010-07-09 15:04:48 +0000
abaec13e38Bug fix: if there are samples in the VCF but all of them are no-calls, we still need to emit GT for the FORMAT field to be on spec. Note that this is a holdover from 3.3 writing but can't easily be fixed there. Fortunately, that code is all going away soon...
ebanks
2010-07-09 14:08:25 +0000
ea8fd506bfUpdate to PickSequenomProbes: Option to ignore mask sites within X bp of a variant (very useful for indels where dbSNP entries near the indel are almost always false SNP calls). Also fixed an integration test where the variant site itself, being in dbSNP, was represented as [N/C] rather than [A/C]. Added integration test for 1bp no-mask window.
chartl
2010-07-09 04:03:19 +0000
179067e3f4Support for . values in qual field
depristo
2010-07-09 01:47:02 +0000
dd978dd525misc. changes to python scripts
depristo
2010-07-09 00:13:35 +0000
6ffcaa0afeCan run R scripts on the command line
depristo
2010-07-09 00:13:18 +0000
45fb614296Fixes to VE for obscure bug, as well as disabled integration test for CombineVariants
depristo
2010-07-09 00:13:07 +0000
5d39cd5db8Added --fdr_filter_level to ApplyVariantCuts so that you can create beautiful tranche plots and also decide which tranche level to filter at. The previous version always filtered at the smallest tranche. The tranche filter names are appropriately added to the VCF header.
rpoplin
2010-07-08 22:44:10 +0000
760aaeda88Update to CombineVariants. Now splits merge options into variant and genotype options separately.
depristo
2010-07-08 20:09:48 +0000
bd2ba3eb37deal with very large known indels that fall off our ref context
ebanks
2010-07-08 20:05:16 +0000
7ff6106c14adding Ivy lines for findbug, and adding a build task (to run it locally you need to have installation of findbug). I'll put more information on the wiki when it's up and running.
aaron
2010-07-08 19:10:19 +0000
12fecc8d8fremove the picard DbSNP ROD.
aaron
2010-07-08 17:46:00 +0000
56a0c7ee6fAll headers are now converted to VCF4 by default.
depristo
2010-07-08 14:14:17 +0000
6e6ad36523reallow MNP events through
ebanks
2010-07-08 06:26:52 +0000
ed0d0d78facorresponding fix for dealing with insertions
ebanks
2010-07-08 05:25:03 +0000
ada8c9931fWe were never clipping the VCF-provided ref base off the left end of the alleles for insertions, so the reference allele was never null (and downstream walkers would fail). Didn't this get tested with insertions at some point?
ebanks
2010-07-08 05:24:27 +0000
9a81f1d7efFixed this tool for chartl so that it now properly handles deletions. Added deletion case to integration tests.
ebanks
2010-07-08 04:45:59 +0000
b7a3d1e61fBug fix: if the FORMAT field consisted of just GT, we were exceptioning out. How did we not catch this until now?
ebanks
2010-07-08 04:41:40 +0000
9fc05ac2aeeagerDecode is now false.
hanna
2010-07-07 22:51:48 +0000
4bc3ad2194Shame on me: UG was emitting negative QUALs (-0) in all_bases mode. Thanks, Matt.
ebanks
2010-07-07 20:30:22 +0000
30714ec8d9As per quick chat with Richard Durban, don't increase the mapping quality of realigned reads too much; for now, arbitrarily increase the MQ by 10. We need to figure out a better solution.
ebanks
2010-07-07 20:12:59 +0000
8ff1a4b929Don't try to clean reads that fail the PF, in preparation for Ryan
ebanks
2010-07-07 19:49:36 +0000
2ac74a862dupdated to deal with new cleaner options
ebanks
2010-07-07 17:06:00 +0000
b934cc7554Updates to fix some bugs in merger. Now able to merge into project wide indel VCF files. Integration teests coming tomorrow
depristo
2010-07-07 03:16:33 +0000
7be8c35eb2Workaround for scala trait erasing parameterized types: - Requiring explicit @ClassType on parameterized fields in traits. - Scatter / Gather functions are now abstract classes since @ClassType can't be used on parameterized fields with type parameters.
kshakir
2010-07-07 03:15:10 +0000
120f90da5bInterval support for ref walkers while streaming.
hanna
2010-07-07 03:14:59 +0000
773a72e6eaAn initial fix for performance issues when filtering UG with new StratifiedAlignmentContext.
hanna
2010-07-07 01:07:46 +0000
be75b087eca) Add input argument (-ncrate) to BeagleOutputToVCFWalker. If the genotype posterior error probability is higher than this threshold, we declare No-call at this genotype. b) Add "OG" annotation to genotypes. If Beagle changes genotypes, this annotation gets the original genotype call, to ease performance comparisons. If not, this annotation gets an empty value.
delangel
2010-07-06 18:33:28 +0000
4213e05aebFix for sharding ref walkers via monolithic sharding. Introduces the potential bug (for monolithic sharding only) that when traversing by read, map() function will not be called for loci off the end of the reference.
hanna
2010-07-06 04:34:38 +0000
86031f4034part two: todo's in combine variants, fixes for InferredGeneticContext, and some other tests and clean-up.
aaron
2010-07-05 21:07:53 +0000
36edc60cccConnected UG to the new comp track annotation system in VA. Also, when emit confidence is lower than call confidence (so that we emit records filtered with LowQual), add a corresponding FILTER header field to the VCF so that the validator doesn't complain.
ebanks
2010-07-05 13:04:24 +0000
3347d1ca7cpart one of combining format and info header lines code into a single abstract class for Mark; plus some 'm' removals from access methods for Eric. Adding fixes for CombineVariants next.
aaron
2010-07-05 05:57:58 +0000
e7220bc885Variant Context simple merging routine should keep ID if one of the VCs has it
ebanks
2010-07-05 01:10:15 +0000
3016e1cf80Fixes to increase robustness in vcf4 writer. We assume that only at most 1 base was clipped from beginning of allele encoding by reader, and improve the way we find if bases were clipped. We still cant deal with some corner cases, and duplicate records may follow, for example if a snp location is followed at the next base by an indel. Also, if we are reading form a 3.3 vcf and the reference is null (ie we have an insertion), the reference base is not computed correctly.
delangel
2010-07-04 20:22:04 +0000
07945040f8Set VariantFiltration's JEXL engine to silent for warning messages
ebanks
2010-07-04 18:11:19 +0000
be8740b00dAnother edge case in left alignment for indels: deal with cases when insertions are ambiguously placed at ends of reads
ebanks
2010-07-04 17:26:38 +0000
cd2e4b0a1emerging now very close to working. Bug todo in writer and vcf infrastructure. Can almost create merged snp and indel files
depristo
2010-07-02 20:09:25 +0000
b6bdd61283a) Fix bug when multi-base reference is homopolymeric when writing a VCF4.0 variant context: computation of number of trailing bases was incorrect and we ended up with incorrect position. b) Updated VCF4WriterTestWalker to take either VCF3 or VCF4 as inputs (this walker can also be used to convert from 3.3 to 4.0).
delangel
2010-07-02 15:19:42 +0000
61e2b2e39bNearly finalize merging capabilities for CombineVariants. Support for dealing with inconsistent indel alleles at loci. Improvements to Allele and removal of addAllele to MutableGenotype. We are close to being able to merge all of 1000 genomes -- snps and indels -- into a single combined vcf
depristo
2010-07-02 13:32:33 +0000
cab8394103The sharding system now buffers reads, with a size determined by command-line argument. Will investigate whether/how this impacts performance on low-pass data and, if it works well, will create a more automatic version of the tool.
hanna
2010-07-01 22:28:55 +0000
3093a20a55fixing VCF header format and info fields so that they propery emit the unbounded count value correctly for vcf4 or vcf3. Eric we should update the vcf4 spec page to indicate format fields are allowed to use the unbounded count as well (if this is true).
aaron
2010-07-01 22:02:16 +0000
61c07c6f90Fixes for missing key values that can create null pointer exceptions when reading from 3.3-generated variant contexts. Also, chop missing genotype fields correctly from right to left
delangel
2010-07-01 20:17:03 +0000
87470d5fe5Checking in a simplistic VR qscript file for posterity's sake
rpoplin
2010-07-01 18:53:17 +0000
255b036fb5Variant Recalibrator MLE EM algorithm is moved over to variational Bayes EM in order to eliminate problems with singularities when clustering in higher than two dimensions. Because of this there is no longer a number of Gaussians parameter. Wiki will be updated shortly with new recommended command.
rpoplin
2010-07-01 18:51:07 +0000
4903d1fb4ffix for a parallelization issue: moving the creation of iterators outside of the sync block so we don't wait for RMD tracks to seek to the correct location. Thanks to Ben for providing the test case!
aaron
2010-07-01 16:37:02 +0000
43ca595d15VCF headers now can be set to a particular VCF version after creation, which converts the header lines to the appropriate encoding on output. Plus some clean-up of the code.
aaron
2010-07-01 05:32:14 +0000
4995950d04IndexedFastaSequenceFile is now in Picard; transitioning to that implementation.
hanna
2010-07-01 04:40:31 +0000
b6af17b82dRev Picard with new IndexedFastaSequenceFile patch.
hanna
2010-07-01 03:05:43 +0000
c9d5345150Redo StratifiedAlignmentContext to use ReadBackedPileup's stratification options.
hanna
2010-07-01 02:46:05 +0000
dc4715c9c6Permit empty fields in INFO and FORMAT structures - not fully tested yet but at least failing cases before now pass. Also, corrected a bug where in case we were reading 3.3 VCF's, or VCFs with no original allele encodings, we'd always print 2 bases per allele.
delangel
2010-07-01 01:56:07 +0000
15a6be5d50Updated to reflect Mark's renaming binge
ebanks
2010-06-30 22:48:42 +0000
5f2b2d860eFinal stage of renaming
depristo
2010-06-30 21:39:07 +0000
6e7927a47dContinuing the renaming nightmare...
depristo
2010-06-30 20:25:01 +0000
9d7d5f1747Continuing the renaming nightmare...
depristo
2010-06-30 20:24:27 +0000
b8d6a95e7aPreliminary commit of new VCFCombine, soon to be called CombineVariants (next commit) that support merging any number of VCF files via a general VC merge routine that support prioritization and merging of samples! It's now possible to merge the pilot1/2/3 call sets into a single (monster) VCF taking genotypes from pilot2, then pilot3, then pilot1 as needed.
depristo
2010-06-30 20:13:03 +0000
a46e22ed13Refactored ArgumentDefinition to absorb functionality from ArgumentDefinition and ArgumentTypeDescriptor.
kshakir
2010-06-30 18:55:57 +0000
178cf64a0cRefactored ArgumentDefinition to absorb functionality from ArgumentDefinition and ArgumentTypeDescriptor.
kshakir
2010-06-30 18:37:58 +0000
569456850dMark pointed out there's differentiation in the filter field. Rolling back.
chartl
2010-06-30 17:05:53 +0000
52a474b27dFixed an issue with VCF combine in sites like the following:
chartl
2010-06-30 16:49:31 +0000
66931d433cuseful routines for R
depristo
2010-06-30 16:38:49 +0000
944dbb94ceRefactored and generalized the database/comp annotations in VariantAnnotator. Now one can provide comp tracks as with VariantEval (e.g. compHapMap, comp1KG_CEU) and the INFO field will be annotated with the track name (without the 'comp') if the variant record overlaps a comp site (e.g. ...;1KG_CEU;...). This means that you can now pass 1kg calls to the Unified Genotyper and automatically have records annotated with their presence in 1kg.
ebanks
2010-06-30 16:37:31 +0000
cf910d9cc2misc. useful updates to python library
depristo
2010-06-30 16:33:32 +0000
47c4a70ac1It turns out that it is legitimately possible for there to be reads that won't overlap within a target interval for cleaning. While we don't want to attempt cleaning, we also don't want to fail.
ebanks
2010-06-30 15:50:44 +0000
ae33d8a2f2I just wanted one more vote. It's settled: we die.
ebanks
2010-06-30 14:00:56 +0000
8fb37f5f7aFor Kiran: warn the user when the actual and vcf ref bases differ so that if an exception is generated later, he knows why. All: should we generate the actual exception here? Is there any reason to allow cases where the vcf record has a different ref base than the actual reference? I'd vote that we die here. Thoughts?
ebanks
2010-06-30 13:56:16 +0000
d932322190More necessary fixes for VCF4.0 - now results look more sensible in realistic, bigger VCF files produced by say Dindel and not just the small test VCF: - Fixed and cleaned code to produce trailing and padding bases in alleles around indels. - Deal better with missing fields. Pending: - Chopping missing fields at end of genotypes.
delangel
2010-06-30 02:59:30 +0000
12c0de6170Added ability to clean using only known indels. Added integration test for it. Fixed vcf->vc conversion for indels which was busted.
ebanks
2010-06-30 01:20:56 +0000
610cc7ae2bCool package trick Kiran showed me. VariantEvaluator no longer public, AAT specifies the core package even though it lives in oneoffs. Disabled so integration tests pass.
chartl
2010-06-29 22:42:04 +0000
4c6f4e41c6Include making VariantEvaluator public within the package so my oneoffs can be seen (not included in previous submit specifically because I didn't want to break the build by changing anything in core...the road to hell is paved with good intentions)
chartl
2010-06-29 22:26:52 +0000
9ac13b8f5dName and body change for this module to reflect local code.
chartl
2010-06-29 21:45:26 +0000
844cb2ed33fixing a bug that Eric found with RODs for reads, where some records could be omitted. Sorry Eric!
aaron
2010-06-29 21:38:55 +0000
101c27294dComment this guy out so we build again. (Hate it when my repository goes all funky.)
chartl
2010-06-29 21:16:33 +0000
3017f82550Initial commit of items for analyzing amino acid transitions in variant eval. Blew up my subversion by coding locally while i did not have internet. I hope this doesn't bust any integrationtests since I changed no existing code but...who knows. Crossing my fingers.
chartl
2010-06-29 20:57:18 +0000
e3fb4d5c70Intermediate checkin, just to fix null pointer exception that happened when merging implementation with latest VCF4 decoder - field ORIGINAL_ALLELE_LIST in vc shouldn't be written in infoFields structure since this won't be output to file and there is no legal structure under this key. Base encoding for complex events is still brittle and most probably still has issues, fixes upcoming.
delangel
2010-06-29 20:57:09 +0000
baf9479c35An addition for Sendu since he can't seem to tell when his CountCovariate jobs die in the middle of writing the CSVs. We now write an EOF marker at the end of the covariates table and look for it when reading in the file in TableRecalibrationWalker. By default, we warn the user if the EOF marker isn't present, but we exception out if the user provides the --fail_with_no_eof_marker option.
ebanks
2010-06-29 18:50:07 +0000
3ca2b7374bFixes to better deal with the "Type" and "Number" field in the INFO and FORMAT header lines in VCF4.0. We now record these fields and provide appropriate conversions. This is the first version that passes fully the VCF validator. Also, moved the flag indicating VCF4.0 to the VCFWriter constructor.
delangel
2010-06-29 16:43:00 +0000
801b47c6e9For Sendu: a similar addition to the Indel Genotyper allowing it to emit a metrics file (which for now consists only of # of normal/tumor calls made)
ebanks
2010-06-29 13:19:17 +0000
ddf87e61c2For Sendu: optionally emit a metrics file with callability info (including number of actual calls made) from UG
ebanks
2010-06-29 12:57:28 +0000
929e5b9276Fix possible null pointer exception
ebanks
2010-06-29 09:01:18 +0000
2953c9f069Efficiency improvement requested by the Picard team in IndexedFastaSequenceFile: improve the memory efficiency (and loading time) of long reference sequences by better controlling the input buffer size.
hanna
2010-06-29 07:22:07 +0000
ed71e53dd41) Initial complete version of VCF4 writer. There are still issues (see below) but at least this version is fully functional. It incorporates getting rid of intermediate VCFRecord so we now operate from VariantContext objects directly to VCF 4.0 output. See VCF4WriterTestWalker for usage example: it just amounts to adding vcfWriter.add(vc,ref.getBases()) in walker.
delangel
2010-06-28 23:54:38 +0000
4a451949baadd parallel option to target creator for masking out reads with bad mates
ebanks
2010-06-28 22:13:25 +0000
dce2c17404Added "-bsubWait" where Queue waits for all the jobs to exit before exiting.
kshakir
2010-06-28 19:52:17 +0000
20f5fdbcf7Changes to MVC to make the the header of its output VCF compliant with spec (give expected # of values for info field annotations)
chartl
2010-06-28 18:33:23 +0000
c047232b18Using picard for bam merging. Properties now propagate to scatter/gather functions.
kshakir
2010-06-28 17:59:09 +0000
62d22ff1aaadding the original allele list to a variant context (as the annotation ORIGINAL_ALLELE_LIST), in the case where the set alleles are the result of clipping. Added tests for both cases.
aaron
2010-06-28 17:23:46 +0000
1292c96e29The cleaner now adds the OC (original cigar) and OS (original alignment start) tags as appropriate to reads that get realigned; this feature can be turned off. Also, improved integration tests (sorry, Kiran!).
ebanks
2010-06-28 16:46:47 +0000
depristo
ebanks
chartl
rpoplin
aaron
kshakir
hanna
delangel
weisburd