* add a new column to do what I have been doing manually for every project, understand why we got no usable coverage in that interval
* add unit tests -- this tool is now public, we need tests.
* slightly better docs -- in an effort to produce better docs for this tool
most people don't care about excessive coverage (unless you're very
particular about your analysis). Therefore the best possible default
value for this is Integer.maxValue so it doesn't get in the way.
Itemized Changes:
* change maximumCoverage threshold to Integer.maxValue
[delivers #57353620]
-- Adding changes to CombineVariants to work with the Reference Model mode of the HaplotypeCaller.
-- Added -combineAnnotations mode to CombineVariants to merge the info field annotations by taking the median
-- Added new StrandBiasBySample genotype annotation for use in computing strand bias from single sample input vcfs
-- Bug fixes to calcGenotypeLikelihoodsOfRefVsAny, used in isActive() as well as the reference model
-- Added active region trimming capabilities to the reference model mode, not perfect yet, turn off with --dontTrimActiveRegions
-- We only realign reads in the reference model if there are non-reference haplotypes, a big time savings
-- We only realign reads in the reference model if the read is informative for a particular haplotype over another
-- GVCF blocks will now track and output the minimum PLs over the block
-- MD5 changes!
-- HC tests: from bug fixes in calcGenotypeLikelihoodsOfRefVsAny
-- GVCF tests: from HC changes above and adding in active region trimming
PairHMMSyntheticBenchmark and PairHMMEmpirical benchmark were written to test the banded pairHMM, and were archived along with it. I returned them to the test directory for use in benchmarking the ArrayLoglessPairHMM. I commented out references to the banded pairHMM (which was left in archive), rather than removing those references entirely.
Renamed PairHMMEmpiricalBenchmark to PairHMMBandedEmpiricalBenchmark and returned it to the archive. It has a few problems for use as a general benchmark, including initializing the HMM too frequently and doing too much setup work in the 'time' method. However, since the size selection and debug printing are useful for testing the banded implementation, I decided to keep it as-is and archive it alongside with the other banded pairHMM classes. I did fix one bug that was causing the selectWorkingData function to return prematurely. As a result, the benchmark was only evaluating 4-40 pairHMM calls instead of the desired "maxRecords".
I wrote a new PairHMMEmpiricalBenchmark that simply works through a list of data, with setup work and hmm-initialization moved to its own function. This involved writing a new data read-in function in PairHMMTestData. The original was not maintaining the input data in order, the end result of which would be an over-estimate of how much caching we are able to do. The new benchmark class more closely mirrors real-world operation over large data.
It might be cleaner to fix some of the issues with the BandedEmpiricalBenchmark and use one read-in function. However, this would involve more extensive changes to:
PairHMMBandedEmpiricalBenchmark
PairHMMTestData
BandedLoglessPairHMMUnitTest
I decided against this as the banded benchmark and unit test are archived.
Returned Logless Caching implementation to the default in all cases. Changing to the array version will await performance benchmarking
Refactored many pieces of functionality in ArrayLoglessPairHMM into their own methods.
A new PairHMM implementation for read/haplotype likelihood calculations. Output is the same as the LOGLESS_CACHING version.
Instead of allocating an entire (read x haplotype) matrix for each HMM state, this version stores sub-computations in 1D arrays. It also accesses intersections of the (read x haplotype) alignment in a different order, proceeding over "diagonals" if we think of the alignment as a matrix.
This implementation makes use of haplotype caching. Because arrays are overwritten, it has to explicitly store mid-process information. Knowing where to capture this info requires us to look ahead at the subsequent haplotype to be analyzed. This necessitated a signature change in the primary method for all pairHMM implementations.
We also had to adjust the classes that employ the pairHMM:
LikelihoodCalculationEngine (used by HaplotypeCaller)
PairHMMIndelErrorModel (used by indel genotyping classes)
Made the array version the default in the HaplotypeCaller and the UnifiedArgumentCollection.
The latter affects classes:
ErrorModel
GeneralPloidyIndelGenotypeLikelihoodsCalculationModel
IndelGenotypeLikelihoodsCalculationModel
... all of which use the pairHMM via PairHMMIndelErrorModel
-This was a dependency of the test suite, but not the GATK proper,
which caused problems when running the test suite on the packaged
GATK jar at release time
-Use GATKVCFUtils.readVCF() instead
* Refactoring implementations of readHeader(LineReader) -> readActualHeader(LineIterator), including nullary implementations where applicable.
* Galvanizing fo generic types.
* Test fixups, mostly to pass around LineIterators instead of LineReaders.
* New rev of tribble, which incorporates a fix that addresses a problem with TribbleIndexedFeatureReader reading a header twice in some instances.
* New rev of sam, to make AbstractIterator visible (was moved from picard -> sam in Tribble API refactor).
There is now a command-line option to set the model to use in the HC.
Incorporated Ryan's current (unmerged) branch in for most of these changes.
Because small changes to the math can have drastic effects, I decided not to let users tweak
the calculations themselves. Instead they can select either NONE, CONSERVATIVE (the default),
or AGGRESSIVE.
Note that any base insertion/deletion qualities from BQSR are still used.
Also, note that the repeat unit x repeat length approach gave very poor results against the KB,
so it is not included as an option here.
- Make -rod required
- Document that contaminationFile is currently not functional with HC
- Document liftover process more clearly
- Document VariantEval combinations of ST and VE that are incompatible
- Added a caveat about using MVLR from HC and UG.
- Added caveat about not using -mte with -nt
- Clarified masking options
- Fixed docs based on Erics comments
-- When provided, this argument causes us to only emit the selected samples into the VCF. No INFO field annotations (AC for example) or other features are modified. It's current primary use is for efficiently evaluating joint calling.
-- Add integration test for onlyEmitSamples
-- The previous approach in VQSR was to build a GMM with the same max. number of Gaussians for the positive and negative models. However, we usually have many more positive sites than negative, so we'd prefer to use a more detailed GMM for the positive model and a less well defined model using few sites for the negative model.
-- Now the maxGaussians argument only applies to the positive model
-- This update builds a GMM for the negative model with a default 4 max gaussians (though this can be controlled via command line parameter)
-- Removes the percentBadVariants argument. The only way to control how many variants are included in the negative model is with minNumBad
-- Reduced the minNumBad argument default to 1000 from 2500
-- Update MD5s for VQSR. md5s changed significantly due to underlying changes in the default GMM model. Only sites with NEGATIVE_TRAINING_LABELs and the resulting VQSLOD are different, as expected.
-- minNumBad is now numBad
-- Plot all negative training points as well, since this significantly changes our view of the GMM PDF
-- In the case where there's some variation to assembly and evaluate but the resulting haplotypes don't result in any called variants, the reference model would exception out with "java.lang.IllegalArgumentException: calledHaplotypes must contain the refHaplotype". Now we detect this case and emit the standard no variation output.
-- [delivers #54625060]
Problem
-------
Caching strategy is incompatible with the current sorting of the haplotypes, and is rendering the cache nearly useless.
Before the PairHMM updates, we realized that a lexicographically sorted list of haplotypes would optimize the use of the cache. This was only true until we've added the initial condition to the first row of the deletion matrix, which depends on the length of the haplotype. Because of that, every time the haplotypes differ in length, the cache has to be wiped. A lexicographic sorting of the haplotypes will put different lengths haplotypes clustered together therefore wasting *tons* of re-compute.
Solution
-------
Very simple. Sort the haplotypes by LENGTH and then in lexicographic order.
1. Removing old legacy code that was capping the positional depth for reduced reads to 127.
Unfortunately this cap affectively performs biased down-sampling and throws off e.g. FS numbers.
Added end to end unit test that depth counts in RR can be higher than max byte.
Some md5s change in the RR tests because depths are now (correctly) no longer capped at 127.
2. Down-sampling in ReduceReads was not safe as it could remove het compressed consensus reads.
Refactored it so that it can only remove non-consensus reads.
Now only filtered reads are unstranded. All consensus reads have strand, so that we
emit 2 consensus reads in general now: one for each strand.
This involved some refactoring of the sliding window which cleaned it up a lot.
Also included is a bug fix:
insertions downstream of a variant region weren't triggering a stop to the compression.
So, compromise solution is to go back to having biallelic PLs but emit a new FORMAT field, called APL, which has the 10 values, but all other statistics and regular PLs are computed as before.
Note that integration test had to be disabled, as the BCF2 codec apparently doesn't support writing into genotype fields other than PL,DP,AD,GQ,FT and GT.
Problem
-------
Qualify Missing Intervals only accepted GATK formatted interval files for it's coding sequence and bait parameters.
Solution
-------
There is no reason for such limitation, I erased all the code that did the parsing and used IntervalUtils to parse it (therefore, now it handles any type of interval file that the GATK can handle).
ps: Also added an average depth column to the output
- Added integration test to show that providing a contamination value and providing same value via a file results in the same VCF
- overrode default contamination value in test
1. Some minor refactorings and claenup (e.g. removing unused imports) throughout.
2. Updates to the KB assessment functionality:
a. Exclude duplicate reads when checking to see whether there's enough coverage to make a call.
b. Lower the threshold on FS for FPs that would easily be filtered since it's only single sample calling.
3. Make the HC consistent in how it treats the pruning factor. As part of this I removed and archived
the DeBruijn assembler.
4. Improvements to the likelihoods for the HC
a. We now include a "tristate" correction in the PairHMM (just like we do with UG). Basically, we need
to divide e by 3 because the observed base could have come from any of the non-observed alleles.
b. We now correct overlapping read pairs. Note that the fragments are not merged (which we know is
dangerous). Rather, the overlapping bases are just down-weighted so that their quals are not more
than Q20 (or more specifically, half of the phred-scaled PCR error rate); mismatching bases are
turned into Q0s for now.
c. We no longer run contamination removal by default in the UG or HC. The exome tends to have real
sites with off kilter allele balances and we occasionally lose them to contamination removal.
5. Improved the dangling tail merging implementation.
-- Assembly graph building now returns an object that describes whether the graph was successfully built and has variation, was succesfully built but didn't have variation, or truly failed in construction. Fixing an annoying bug where you'd prefectly assembly the sequence into the reference graph, but then return a null graph because of this, and you'd increase your kmer because it null was also used to indicate assembly failure
--
-- Output format looks like:
20 10026072 . T <NON_REF> . . . GT:AD:DP:GQ:PL 0/0:3,0:3:9:0,9,120
20 10026073 . A <NON_REF> . . . GT:AD:DP:GQ:PL 0/0:3,0:3:9:0,9,119
20 10026074 . T <NON_REF> . . . GT:AD:DP:GQ:PL 0/0:3,0:3:9:0,9,121
20 10026075 . T <NON_REF> . . . GT:AD:DP:GQ:PL 0/0:3,0:3:9:0,9,119
20 10026076 . T <NON_REF> . . . GT:AD:DP:GQ:PL 0/0:3,0:3:9:0,9,120
20 10026077 . T <NON_REF> . . . GT:AD:DP:GQ:PL 0/0:3,0:3:9:0,9,120
20 10026078 . C <NON_REF> . . . GT:AD:DP:GQ:PL 0/0:5,0:5:15:0,15,217
20 10026079 . A <NON_REF> . . . GT:AD:DP:GQ:PL 0/0:6,0:6:18:0,18,240
20 10026080 . G <NON_REF> . . . GT:AD:DP:GQ:PL 0/0:6,0:6:18:0,18,268
20 10026081 . T <NON_REF> . . . GT:AD:DP:GQ:PL 0/0:7,0:7:21:0,21,267
We use a symbolic allele to indicate that the site is hom-ref, and because we have an ALT allele we can provide AD and PL field values. Currently these are calculated as ref vs. any non-ref value (mismatch or insertion) but doesn't yet account properly for alignment uncertainty.
-- Can we enabled for single samples with --emitRefConfidence (-ERC).
-- This is accomplished by realigning the each read to its most likley haplotype, and then evaluting the resulting pileups over the active region interval. The realignment is done by the HaplotypeBAMWriter, which now has a generalized interface that lets us provide a ReadDestination object so we can capture the realigned reads
-- Provide access to the more raw LocusIteratorByState constructor so we can more easily make them programmatically without constructing lots of misc. GATK data structures. Moved the NO_DOWNSAMPLING constant from LIBSDownsamplingInfo to LocusIteratorByState so clients can use it without making LIBSDownsamplingInfo a public class.
-- Includes GVCF writer
-- Add 1 mb of WEx data to private/testdata
-- Integration tests for reference model output for WGS and WEx data
-- Emit GQ block information into VCF header for GVCF mode
-- OutputMode from StandardCallerArgumentCollection moved to UnifiedArgumentCollection as its no longer relevant for HC
-- Control max indel size for the reference confidence model from the command line. Increase default to 10
-- Don't use out_mode in HaplotypeCallerComplexAndSymbolicVariantsIntegrationTest
-- Unittests for ReferenceConfidenceModel
-- Unittests for new MathUtils functions
-- The previous code would adapter clip before reverting soft clips, so because we only clip the adapter when it's actually aligned (i.e., not in the soft clips) we were actually not removing bases in the adapter unless at least 1 bp of the adapter was aligned to the reference. Terrible.
-- Removed the broken logic of determining whether a read adaptor is too long.
-- Doesn't require isProperPairFlag to be set for a read to be adapter clipped
-- Update integration tests for new adapter clipping code
I "fixed" this once before but instead of testing with unit tests I used integration tests.
Bad decision.
The proper fix is in now, with a bonafide unit test included.
Previous fixes and tests only covered trailing soft-clips. Now that up front
hard-clipping is working properly though, we were failing on those in the tool.
Added a patch for this as well as a separate test independent of the soft-clips
to make sure that it's working properly.
This time we don't accidentally drop reads (phew), but this bug does cause us not to
update the alignment start of the mate. Fixed and added unit test to cover it.
-- Added experimental LikelihoodRankSum, which required slightly more detailed access to the information managed by the base class, so added an overloaded getElementForRead also provides access to the MostLikelyAllele class
-- Added base class default implementation of getElementForPileupElement() which returns null, indicating that the pileup version isn't supported.
-- Added @Override to many of the RankSum classes for safety's sake
-- Updates to GeneralCallingPipeline: annotate sites with dbSNP IDs,
-- R script to assess the value of annotations for VQSR
-- The VR, when the model is bad, may evaluate log10sumlog10 where some of the values in the vector are NaN. This case is now trapped in VR and handled as previously -- indicating that the model has failed and evaluation continues.
-- Currently we don't support writing a BAM file from the haplotype caller when nct is enabled. Check in initialize if this is the case, and throw a UserException
-- Previous version emitted command lines that look like:
##HaplotypeCaller="analysis_type=HaplotypeCaller input_file=[private/testdata/reduced.readNotFullySpanningDeletion.bam] ..."
the new version provides additional information on when the GATK was run and the GATK version in a nicer format:
##GATKCommandLine=<ID=HaplotypeCaller,Version=2.5-206-gbc7be2b,Date="Thu Jun 20 11:09:01 EDT 2013",Epoch=1371740941197,CommandLineOptions="analysis_type=HaplotypeCaller input_file=[private/testdata/reduced.readNotFullySpanningDeletion.bam] read_buffer_size=null phone_home=AWS ...">
-- Additionally, the command line options are emitted sequentially in the file, so you can see a running record of how a VCF was produced, such as this example from the integration test:
##GATKCommandLine=<ID=HaplotypeCaller,Version=2.5-206-gbc7be2b,Date="Thu Jun 20 11:09:01 EDT 2013",Epoch=1371740941197,CommandLineOptions="lots of stuff">
##GATKCommandLine=<ID=SelectVariants,Version=2.5-206-gbc7be2b,Date="Thu Jun 20 11:16:23 EDT 2013",Epoch=1371741383277,CommandLineOptions="lots of stuff">
-- Removed the ProtectedEngineFeaturesIntegrationTest
-- Actual unit tests for these features!
Improved AnalyzeCovariates (AC) integration test.
Renamed AC test files ending with .grp to .table
Implementation:
* Removed RECAL_PDF/CSV_FILE from RecalibrationArgumentCollection (RAC). Updated rest of the code accordingly.
* Fixed BQSRIntegrationTest to work with new changes
Implemtation details:
* Added tool class *.AnalyzeCovariates
* Added convenient addAll method to Utils to be able to add elements of an array.
* Added parameter comparison methods to RecalibrationArgumentCollection class in order to verify that multiple imput recalibration report are compatible and comparable.
* Modified the BQSR.R script to handle up to 3 different recalibration tables (-BQSR, -before and -after) and removed some irrelevant arguments (or argument values) from the output.
* Added an integration test class.
-- Changed default HMM model.
-- Removed check.
-- Changed md5's: PL's in the high 100s change by a point or two due to new implementation.
-- Resulting performance improvement is about 30 to 50% less runtime when using -glm INDEL.
-- numPruningSamples allows one to specify that the minPruning factor must be met by this many samples for a path to be considered good (e.g. seen twice in three samples). By default this is just one sample.
-- adding unit test to test this new functionality
-- When doing cross-species comparisons and studying population history and ancient DNA data, having SOME measure of confidence is needed at every single site that doesn't depend on the reference base, even in a naive per-site SNP mode. Old versions of GATK provided GQ and some wrong PL values at reference sites but these were wrong. This commit addresses this need by adding a new UG command line argument, -allSitePLs, that, if enabled will:
a) Emit all 3 ALT snp alleles in the ALT column.
b) Emit all corresponding 10 PL values.
It's up to the user to process these PL values downstream to make sense of these. Note that, in order to follow VCF spec, the QUAL field in a reference call when there are non-null ALT alleles present will be zero, so QUAL will be useless and filtering will need to be done based on other fields.
-- Tweaks and fixes to processing pipelines for Reich lab.
1. Have the RMSMappingQuality annotation take into account the fact that reduced reads represent multiple reads.
2. The rank sume tests should not be using reduced reads (because they do not represent distinct observations).
3. Fixed a massive bug in the BaseQualityRankSumTest annotation! It was not using the base qualities but rather
the read likelihoods?!
Added a unit test for Rank Sum Tests to prove that the distributions are correctly getting assigned appropriate p-values.
Also, and just as importantly, the test shows that using reduced reads in the rank sum tests skews the results and
makes insignificant distributions look significant (so it can falsely cause the filtering of good sites).
Also included in this commit is a massive refactor of the RankSumTest class as requested by the reviewer.
-- Previous version created FILTERs for each possible alt allele when that site was set to monomorphic by BEAGLE. So if you had a A/C SNP in the original file and beagle thought it was AC=0, then you'd get a record with BGL_RM_WAS_A in the FILTER field. This obviously would cause problems for indels, as so the tool was blowing up in this case. Now beagle sets the filter field to BGL_SET_TO_MONOMORPHIC and sets the info field annotation OriginalAltAllele to A instead. This works in general with any type of allele.
-- Here's an example output line from the previous and current versions:
old: 20 64150 rs7274499 C . 3041.68 BGL_RM_WAS_A AN=566;DB;DP=1069;Dels=0.00;HRun=0;HaplotypeScore=238.33;LOD=3.5783;MQ=83.74;MQ0=0;NumGenotypesChanged=1;OQ=1949.35;QD=10.95;SB=-6918.88
new: 20 64062 . G . 100.39 BGL_SET_TO_MONOMORPHIC AN=566;DP=1108;Dels=0.00;HRun=2;HaplotypeScore=221.59;LOD=-0.5051;MQ=85.69;MQ0=0;NumGenotypesChanged=1;OQ=189.66;OriginalAltAllele=A;QD=15.81;SB=-6087.15
-- update MD5s to reflect these changes
-- [delivers #50847721]
-- Now table looks like:
Name VariantType AssessmentType Count
variant SNPS TRUE_POSITIVE 1220
variant SNPS FALSE_POSITIVE 0
variant SNPS FALSE_NEGATIVE 1
variant SNPS TRUE_NEGATIVE 150
variant SNPS CALLED_NOT_IN_DB_AT_ALL 0
variant SNPS HET_CONCORDANCE 100.00
variant SNPS HOMVAR_CONCORDANCE 99.63
variant INDELS TRUE_POSITIVE 273
variant INDELS FALSE_POSITIVE 0
variant INDELS FALSE_NEGATIVE 15
variant INDELS TRUE_NEGATIVE 79
variant INDELS CALLED_NOT_IN_DB_AT_ALL 2
variant INDELS HET_CONCORDANCE 98.67
variant INDELS HOMVAR_CONCORDANCE 89.58
-- Rewrite / refactored parts of subsetDiploidAlleles in GATKVariantContextUtils to have a BEST_MATCH assignment method that does it's best to simply match the genotype after subsetting to a set of alleles. So if the original GT was A/B and you subset to A/B it remains A/B but if you subset to A/C you get A/A. This means that het-alt B/C genotypes become A/B and A/C when subsetting to bi-allelics which is the convention in the KB. Add lots of unit tests for this functions (from 0 previously)
-- BadSites in Assessment now emits TP sites with discordant genotypes with the type GENOTYPE_DISCORDANCE and tags the expected genotype in the info field as ExpectedGenotype, such as this record:
20 10769255 . A ATGTG 165.73 . ExpectedGenotype=HOM_VAR;SupportingCallsets=ebanks,depristo,CEUTrio_best_practices;WHY=GENOTYPE_DISCORDANCE GT:AD:DP:GQ:PL 0/1:1,9:10:6:360,0,6
Indicating that the call was a HET but the expected result was HOM_VAR
-- Forbid subsetting of diploid genotypes to just a single allele.
-- Added subsetToRef as a separate specific function. Use that in the DiploidExactAFCalc in the case that you need to reduce yourself to ref only. Preserves DP in the genotype field when this is possible, so a few integration tests have changed for the UG
-- Merging overlapping fragments turns out to be a bad idea. In the case where you can safely merge the reads you only gain a small about of overlapping kmers, so the potential gains are relatively small. That's in contrast to the very large danger of merging reads inappropriately, such as when the reads only overlap in a repetitive region, and you artificially construct reads that look like the reference but actually may carry a larger true insertion w.r.t. the reference. Because this problem isn't limited to repetitive sequeuence, but in principle could occur in any sequence, it's just not safe to do this merging. Best to leave haplotype construction to the assembly graph.
We now run Smith-Waterman on the dangling tail against the corresponding reference tail.
If we can generate a reasonable, low entropy alignment then we trigger the merge to the
reference path; otherwise we abort. Also, we put in a check for low-complexity of graphs
and don't let those pass through.
Added tests for this implementation that checks exact SW results and correct edges added.
Principle is simple: when coverage is deep enough, any single-base read error will look like a rare k-mer but correct sequence will be supported by many reads to correct sequences will look like common k-mers. So, algorithm has 3 main steps:
1. K-mer graph buildup.
For each read in an active region, a map from k-mers to the number of times they have been seen is built.
2. Building correction map.
All "rare" k-mers that are sparse (by default, seen only once), get mapped to k-mers that are good (by default, seen at least 20 times but this is a CL argument), and that lie within a given Hamming distance (by default, =1). This map can be empty (i.e. k-mers can be uncorrectable).
3. Correction proposal
For each constituent k-mer of each read, if this k-mer is rare and maps to a good k-mer, get differing base positions in k-mer and add these to a list of corrections for each base in each read. Then, correct read at positions where correction proposal is unanimous and non-empty.
The algorithm defaults are chosen to be very stringent and conservative in the correction: we only try to correct singleton k-mers, we only look for good k-mers lying at Hamming distance = 1 from them, and we only correct a base in read if all correction proposals are congruent.
By default, algorithm is disabled but can be enabled in HaplotypeCaller via the -readErrorCorrect CL option. However, at this point it's about 3x-10x more expensive so it needs to be optimized if it's to be used.
Geraldine Van der Auwera
Mauricio Carneiro
Ryan Poplin
Eric Banks
Louis Bergelson
bradtaylor
David Roazen
Michael McCowan
Mark DePristo
Yossi Farjoun
Joseph Rose
sathibault
Guillermo del Angel
Scott Thibault
Scott Thibault
Valentin Ruano-Rubio