- Brought over exact AF estimation from branch (which is now dead). Exact model is default in UnifiedGenotyperV2.
- Implemented completely new genotyping algorithm given best AF estimate using dynamic programming, which in theory should be better than both greedy search and any HWE-based genotyper.
- Integrated and added new Dindel likelihood estimation model.
- Corrected annotators that would call readBasePileup: since we can be annotating extended events, best way is to interrogate context for kind of pileup and either readBasePileup or readExtendedEventPileup.
All changes above except last one are still in playground since they require more testing.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4396 348d0f76-0448-11de-a6fe-93d51630548a
pieces of core that depend on playground. Most of these have been eliminated by
(temporarily) promoting Aaron's report system to core in this checkin. I'll
follow up with other changes in separately.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4350 348d0f76-0448-11de-a6fe-93d51630548a
a) Redid way to compute path metrics in indel error model. Paper formulation where we have an anchor point in the alignemt between read and haplotype won't work in practice except in nice data sets that are perfectly indel-realigned and that are well mapped by aligner. New formulation doesn't assume this, and it's actually simpler and uses less code. It now resembles more a classic SW dynamic programming formulation but it still preserves the HMM probabilistic formulation.
b) Added a programmable call threshold, set by command line.
c) Use now sample name from BAM file, remove -sampleName argument.
d) Simplify loop to compute read-haplotype likelihoods.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4311 348d0f76-0448-11de-a6fe-93d51630548a
*** Three integration tests had to change: ***
RecalibarationWalkersIntegrationTest:
One of the tests was using the interval as the snp track, and wasn't supplying a DbSNP track (for CountCovariates)
SequenomValidationConverterIntegrationTest:
relies on Plink ROD which we've removed.
PileupWalkerIntegrationTest:
we no longer have implicit interval tracks, so there isn't a rod name over the specified region. Otherwise the same result.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4292 348d0f76-0448-11de-a6fe-93d51630548a
a) Turns out previous change of centering haplotype around indel was a bad idea. Context to the left of indel is important but not as important as right one, because by definition all alleles start at the same location, so haplotype is the same to the left of indel regardless of allele. So, go back to having a constant size window to the left of event.
b) Expand reference context so we can test larger haplotypes.
c) Optimize computation of read likelihoods by doing them in linear array instead of in a matrix - no difference in biallelic sites but could be significantly faster in multiallelic sites.
d) Bug fix: read alignment wasn't being computed correctly if, a) we were at an insertion, b) read started right at the insertion, c) read CIGAR didn't include insertion - more of these corner conditions are lurking, so a revamped computation of how reads align to candidate haplotypes is in the works.
e) Add debug option not to use prior haplotype likelihoods.
f) Don't hard-code NA12878 for genotyping, now sample name is a required input argument.
g) Bug fix: if there are no reads covering a candidate indel event, just output NO_CALL (didn't notice this in HiSeq, but in P1 data it happens all the time). I need to add a confidence threshold for calling later on.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4291 348d0f76-0448-11de-a6fe-93d51630548a
- Add a simple calculation model for Pr(R|H) that doesn't rely on Dindel's HMM model. MUCH faster, at a cost of slightly worse performance since we're more sensitive to bad reads coming from sequencing artifacts (add -simple to command line to activate).
- Add debug option to calculation model so that we can optionally output useful info on current read being evaluated. (add -debugout to commandline).
- Small performance improvement: instead of evaluating haplotype to the right of indel (just with a 5 base addition to the left), it seems better to center the indel and to add context evenly around event.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4257 348d0f76-0448-11de-a6fe-93d51630548a
a) Ability to specify haplotype size from command line
b) Expand reference context window so we can form haplotypes for longer indel events.
c) small bug fix in temp output writer (to be removed once I can emit vcfs)
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4212 348d0f76-0448-11de-a6fe-93d51630548a
- GATKVCFWriter deleted, to be replaced if absolutely necessary when VCF writing goes into Tribble.
- VCFWriter is now an interface, for easier redirection.
- VCFWriterImpl fleshes out the VCFWriter interface.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4026 348d0f76-0448-11de-a6fe-93d51630548a
2) Make sure variants are biallelic before asking for isTransversion()
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@4016 348d0f76-0448-11de-a6fe-93d51630548a
- Tribble is included directly in the GATK repo; those who have access to commit to Tribble can now directly commit from the GATK directory from Intellij; command line users can commit from
inside the tribble directory.
- Hapmap ROD now in Tribble; all mentions have been switched over.
- VariantContext does not know about GenomeLoc; use VariantContextUtils.getLocation(VariantContext vc) to get a genome loc.
- VariantContext.getSNPSubstitutionType is now in VariantContextUtils.
- This does not include the checked-in project files for Intellij; still running into issues with changes to the iml files being marked as changes by SVN
I'll send out an email to GSAMembers with some more details.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3954 348d0f76-0448-11de-a6fe-93d51630548a
BTW, Eric, thanks for forwarding the DepthOfCoverage thread to gsamembers. I'd forgotten about reduce by interval. Mighty helpful in this case!
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3906 348d0f76-0448-11de-a6fe-93d51630548a
2. Added some useful logging messages.
3. Added a oneoffs walker to calculate the number of realigned reads and intervals containing them.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3860 348d0f76-0448-11de-a6fe-93d51630548a
b) Bug fixes and update to how we represent indels and other complex events in a VariantContext object. Convention is now that all events are left aligned, with the first variant context location marking the common base before an event occurs. However, alleles in a VC don't have the common base in all VC's. Two new functions are now part of VariantContextUtils: CreateVariantContextWithPaddedAlleles and CreateVariantContextWithTrimmedAlleles. Both take a VC as an input and create a VC as an output.
Main flow is that a VCF reader would create a VC with trimmed alleles, all walkers would ideally work with these trimmed alleles, and then the VCF writer would pad back the alleles before writing. However, there are special cases where we need to pad alleles like for example when merging/combining VC's.
Pending issues:
- PED and DBSNP RODs have to be updated to create VC's for indels following the convention above. Changes will go in after Tribble location is moved and things are tested.
- Need to verify Indel genotyper and other modules that create VC's with indels.- Wiki page describing convention above and how walkers should interpret indel VC's still needs updating/detailing.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3850 348d0f76-0448-11de-a6fe-93d51630548a
2) Keep track of whether vcf records are unfiltered vs. pass filters in the variant context so we can regenerate the records on output.
3) No more "ID" hard-coded all over the code to set the VariantContext ID. Use a static variable instead.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3840 348d0f76-0448-11de-a6fe-93d51630548a
- changed to a better method for getting headers from Codecs
- some removal of old commented out code in the GATKAgrumentCollection
- changes for the rename of FeatureReader to FeatureSource
- removed the old Beagle ROD
- cleaned up some of the code in SampleUtils
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3826 348d0f76-0448-11de-a6fe-93d51630548a
a) VCF track name can work again with 3.3 or 4.0 VCF's when specifying -B name,VCF,file. Code will read header and parse automatically the version.
b) Old VCF codec is deprecated. Reader goes now direct from parsing VCF lines into producing VariantContext objects, with no intermediate VCF records. If anyone can't resist the urge to still input files using the old method, a new VCF3Codec is in place with the old code, but it will be eventually deleted.
c) VCF headers and VCF info fields no longer keep track of the version. They are parsed into an internal representation and will be output only in VCF4.0 format.
d) As a consequence, the existing GATK bug where files are produced with VCF4 body but VCF3.3 headers is solved.
e) Several VCF 4.0 writer bugs are now solved.
f) Integration test MD5's are changed, mostly because of corrected VCF4.0 headers and because validation data mostly uses now VCF4.0.
g) Several VCF files in the ValidationData/ directory have been converted to VCF 4.0 format. I kept the old versions, and the new versions have a .vcf4 extension.
Pending issues:
a) We are still not dealing with indels consistently or correctly when representing them. This will be a second part of the changes.
b) The VCF writer doesn't use VCFRecord but it does still use a lot of leftovers like VCFGenotypeEncoding, VCFGenotypeRecord, etc. This needs to be simplified and cleaned.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3813 348d0f76-0448-11de-a6fe-93d51630548a
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).
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3711 348d0f76-0448-11de-a6fe-93d51630548a
Also putting more tolerance into the timing on the tibble index tests (that check to make sure we're deleting out of date indexes, and not deleting perfectly good indexes). It seems that some of the farm nodes aren't great with a stopwatch.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3674 348d0f76-0448-11de-a6fe-93d51630548a
Also, moved the flag indicating VCF4.0 to the VCFWriter constructor.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3669 348d0f76-0448-11de-a6fe-93d51630548a
See VCF4WriterTestWalker for usage example: it just amounts to adding
vcfWriter.add(vc,ref.getBases()) in walker.
add() method in VCFWriter is polymorphic and can also take a VCFRecord, lthough eventually this should be obsolete.
addRecord is still supported so all backward compatibility is maintained.
Resulting VCF4.0 are still not perfect, so additional changes are in progress. Specifically:
a) INFO codes of length 0 (e.g. HM, DB) are not emitted correctly (they should emit just "HM" but now they emit "HM=1").
b) Genotype values that are specified as Integer in header are ignored in type and are printed out as Doubles.
Both issues should be corrected with better header parsing.
2) Check in ability of Beagle to mask an additional percentage of genotype likelihoods (0 by default), for testing purposes.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3664 348d0f76-0448-11de-a6fe-93d51630548a
1) Some improvements to the VCF4 parsing, including disabling validation.
2) Reimplemented RefSeq in the new Tribble-style rod system.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3630 348d0f76-0448-11de-a6fe-93d51630548a
This check-in will temperarly break the build (I need to see if Bamboo is correctly returning the log file for the failed builds).
Will be fixed once Bamboo starts building.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3609 348d0f76-0448-11de-a6fe-93d51630548a
Provides a cleaner interface with extended events inheriting all of the basic RBP
functionality. Implementation is still slightly messy, but should allow users to
provide separate implementations of methods for sample split pileups and unsplit
pileups for efficiency's sake.
Methods not covered by unit/integration tests have not been sufficiently tested yet.
Unit tests will follow this week.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3597 348d0f76-0448-11de-a6fe-93d51630548a
1. VA is now a ROD walker so it no longer requires reads (needs a little more testing)
2. Annotations can now represent multiple INFO fields (i.e. sets of key/value pairs)
3. The chromosome count annotations have been pulled out of UG and the VCF writer code and into VA where they belong. Fixed the headers too.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3513 348d0f76-0448-11de-a6fe-93d51630548a
+ Identifies opposite homozygote sites
+ Identifies the parent from whom it is expected that a null allele was inherited (or whether it was a putative genotype error; e.g. mom=homref, dad=homref, child=homvar)
+ Labels each opposite homozygote with its homozygous region in the child (e.g. region 1, region 2)
+ Labels each opposite homozygote with the size of the homozygous region in which it was found, the number of child homozygotes in the region, and the number of opposite homozygote violations within that region
To come:
+ Classification of sites as likely tri-allelic
Note that this is very experimental
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3498 348d0f76-0448-11de-a6fe-93d51630548a
More doc/info to follow shortly. Issues still to be solved:
a) Walker changes all genotypes based on Beagle data, but annotations on the original VCF are unchanged. They should in theory be recomputed based on new genotypes.
b) Current implementation is ugly, dirty unwieldy and will necessitate a refactoring soon so I can keep my pride. Most aesthetically affronting issue right now is that we read the full Beagle files at initialization and keep them in memory, but a more delicate implementation would just read from files on a marker by marker basis. Issue that currently prevents this is that BufferedReader() instances don't seem to play nice when called from the map() function.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3488 348d0f76-0448-11de-a6fe-93d51630548a
priors were flat, but fixed nonetheless.
Also, needed to update Tribble.
Minor updates to the Beagle input maker.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3461 348d0f76-0448-11de-a6fe-93d51630548a
@Requires(value={},referenceMetaData=@RMD(name="eval",type= VCFCodec.class))
you'd say:
@Requires(value={},referenceMetaData=@RMD(name="eval",type= VCFRecord.class))
Which is more in-line with what was done before. All instances in the existing codebase should be switched over.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3457 348d0f76-0448-11de-a6fe-93d51630548a
- For speedup in large number of samples, base counts are done on a per read group level, then
merged into counts on larger partitions (samples, libraries, etc)
+ passed all integration tests before next item
- Added additional summary item, a coverage threshold. Set by (possibly multiple) -ct flags,
the summary outputs will have columns for "%_bases_covered_to_X"; both per sample, and
per sample per interval summary files are effected (thus md5s changed for these)
NOTE:
This is the last revision that will include the per-gene summary files. Once DesignFileGenerator is sufficiently general, and has integration tests, it will be moved to core and the per-gene summary from Depth of Coverage will be retired.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3437 348d0f76-0448-11de-a6fe-93d51630548a
passed in as a Tribble SAM text feature. If the generated pileup contains a valid set of reads according to
the downsampling rules, the test passes.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3421 348d0f76-0448-11de-a6fe-93d51630548a
asivache
delangel
hanna
ebanks
aaron
depristo
kiran
chartl
rpoplin
weisburd