* 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.
Eric Banks
Mauricio Carneiro
Geraldine Van der Auwera
Ryan Poplin
Louis Bergelson
bradtaylor
David Roazen
Michael McCowan
Mark DePristo
Yossi Farjoun
Joseph Rose
sathibault
Guillermo del Angel