8fbc0c8473fix for bug GSA-234: fasta index files couldn't handle anything but letters, numbers, or spaces in the contig name
aaron
2009-11-24 19:19:47 +0000
3fca23cd16Added a stub treeReduce function for debugging multi-threaded execution.
andrewk
2009-11-24 18:51:19 +0000
277e6d6b32Further optimizations of TableRecalibration. This completes my goal of having the only math done in the map function be addition, subtraction and rounding the quality score to an integer. Everything else has been moved to the initialize method and only done once.
rpoplin
2009-11-24 18:21:57 +0000
e4546f802cAccumulates coverage across hybrid selection bait intervals to assess effect of bait adjacency. Requires input bait intervals that have an overhang beyond the actual bait interval to capture coverage data at these points. Outputs R parseable file that has all data in lists and then does some basic plotting.
andrewk
2009-11-24 18:12:34 +0000
e5106c9924Hybrid selection performance statistics now include counts of the number of adjacent baits (0,1,2) using OverlapDetector and optionally include assayed bait quantities input via interval lists.
andrewk
2009-11-24 18:07:23 +0000
87c1860398I'm not sure I believe it, but JProfiler claims that calling FourBaseProbs.isVerbose() was taking 5% of my runtime...
ebanks
2009-11-24 17:00:32 +0000
b3f561710fOptimizations: 1. Only do calculations in UG for alternate allele with highest sum of quality scores (note that this also constitutes a bug fix for a precision problem we were having). 2. Avoid using Strings in DiploidGenotype when we can (it was taking 1.5% of my compute according to JProfiler)
ebanks
2009-11-24 16:27:39 +0000
a59e5b5e1aAdded dbSNP sanity check to CountCovariates. If the mismatch rate is too low at dbSNP sites it warns the user that the dbSNP file is suspicious. Added option in CountCovariates and TableRecalibration to ignore read group id's and collapse them together. Also, If the read group is null the walkers no long crash with NullPointerException but instead warn the user the read group and platform are defaulting to some values. Default window size in MinimumNQSCovariate is 5 (two bases in either direction) based on rereading of Chris's analysis.
rpoplin
2009-11-24 16:16:44 +0000
bc35a34f60More informative printing, no longer prints tons of NaN warnings
depristo
2009-11-24 15:45:48 +0000
e5e6d515c3Fix misunderstanding of GenomeLoc interval
alecw
2009-11-24 15:12:49 +0000
cb6d6f2686Very minor performance improvements
ebanks
2009-11-24 05:21:07 +0000
c90bea39a1read.getReadString().charAt(offset) --> read.getReadBases()[offset] [As a courtesy I fixed all instances once I was updating GenotypeLikelihoods]
ebanks
2009-11-24 04:25:19 +0000
ec321abd7bAdded ability to filter on the QUAL field
ebanks
2009-11-24 04:08:22 +0000
36d493e645All standard annotations now inherit from StandardVariantAnnotation. Users can specify whether they want all annotations, just the standard annotations, or specific annotations. When calling in from another walker, the default is just the standard ones.
ebanks
2009-11-24 03:55:12 +0000
ee5093d2c6-Added VariantFiltration integration tests -Added integration test for GLFs
ebanks
2009-11-24 02:36:27 +0000
be6a549e7bAdded the capability to allow expressions in an integration test command (i.e. -filter 'foo') by escaping them in the command.
ebanks
2009-11-24 02:34:48 +0000
903342745dBasic integration test for the aligner.
hanna
2009-11-23 23:08:05 +0000
4837fe919cConvenience changes. If no -BWT option is specified, pull the BWT location from the reference.
hanna
2009-11-23 22:46:05 +0000
9e4eadc37cCountCovariates v2.0.2: Added a --process_nth_locus <int> argument to only use every Nth covered locus when creating the recalibration table.
rpoplin
2009-11-23 22:07:38 +0000
6a52ca3db6Update to the UG integration test. Why I had to rm -rf my entire sting directory to get it to correctly fail we may never know.
chartl
2009-11-23 21:23:00 +0000
ed4cf3de57Check that we're biallelic before calling isSNP()
ebanks
2009-11-23 20:20:48 +0000
5744a1d968The covariates don't care about SAMRecord's anymore - Cleaning up the import statements.
rpoplin
2009-11-23 20:10:12 +0000
af9fe78f84Add a Unix-standard version number to the file.
hanna
2009-11-23 20:07:17 +0000
70059a0fc9Refactored joint estimation model to allow subclasses to overload PofD calculation over all frequencies. Pooled model now takes only 20% of time that it used to. Added integration test for pooled model and updated other joint estimation tests to be more comprehensive now that they are faster.
ebanks
2009-11-23 20:03:38 +0000
7f947f6b60Updated recalibrator integration tests to use all three platforms as well as a bam with multi-platform reads intermingled. CountCovariates v2.0.1: Once again uses a read filter to filter out zero mapping quality reads. Added --sorted_output option to output the table recalibration file in sorted order
rpoplin
2009-11-23 19:51:36 +0000
c299ca5f49It would help if I copied the MD5s from the right integration test... I hate Mondays.
ebanks
2009-11-23 17:21:36 +0000
ff4797acbbForgot to check in integration test update
ebanks
2009-11-23 17:13:51 +0000
14bf6ce83c1. Newest version of the joint estimation model. Faster than previous version and now qscores can get to be > 39.8 for hets. 2. More sanity checks in annotations
ebanks
2009-11-23 17:05:50 +0000
ee2abd30c4Count the best alignments and emit them to a file.
hanna
2009-11-23 16:37:59 +0000
1d46de6d34The old recalibrator is replaced with the refactored recalibrator. Added a version message to the logger output. These walkers start at version 2.0.0
rpoplin
2009-11-23 14:58:33 +0000
dfe7d694711. VCF: don't print slod if it's never set 2. UG: don't print slod if lods are infinite (todo: figure out a good guess instead) 3. UG: if probF=0 for 2 alt alleles are both 0 (because of precision), use log values to discriminate
ebanks
2009-11-23 02:55:43 +0000
753cb100a3Add checks for weird situations
ebanks
2009-11-23 02:14:25 +0000
04d6ac940cAlways print out VCF header - not just when there is genotype data present.
ebanks
2009-11-23 01:44:10 +0000
bf935a6ab11. Fixed bug in PrimaryBaseSecondaryBaseSymmetry code (not checking for null before trying to access object's methods) which was causing Integration Tests to fail. 2. Retired allele frequency range from UG, which wasn't very useful.
ebanks
2009-11-23 01:31:48 +0000
b24240664fReduced the number of calls to new ArrayList() in TableRecalibration. This results in a speed up of perhaps up to 6 percent (timed trials are hard).
rpoplin
2009-11-22 17:24:31 +0000
c9c4999354BWA: odds and ends. Get rid of some spurious debug code that was accidentally checked in. Add a better way to write out unmapped reads (thanks Kiran!) Add a pre-built version of the shared library to the repository for early adoption.
hanna
2009-11-22 15:26:07 +0000
59416ae06aThis is an annotation adapted from one that Mark Daly suggested some time ago. Right now it calculates:
chartl
2009-11-22 00:47:49 +0000
98f921fe24The refactored CountCovariates now hashes the read object into a HashMap which holds all the properties the covariates pull out of the read over and over again such as read group string, bases string and its complement string, quality scores, etc. This results in a big speed up. CountCovariatesRefactored is now just slightly slower than CountCovariates (perhaps 1.07x according to my latest time trial). Thanks to Alec for suggesting IdentityHashMap. CycleCovariate now warns the user that is is defaulting to the Solexa definition of cycle when the platform string pulled out of the read is unrecognized instead of halting with an Exception.
rpoplin
2009-11-21 20:38:17 +0000
27122f7f97Performance improvements for pooled caller. Now possible to actually run on real data in a finite amount of time. Minor changes to GL interface (making strandIndex public) to support cached calculations in pooled caller.
depristo
2009-11-21 15:07:40 +0000
da7de9960bGeneral bug fixes for snpSelector. More robust error checking and handling of NaN values.
depristo
2009-11-21 14:48:29 +0000
797bb83209New VariantFiltration. Wiki docs are updated.
ebanks
2009-11-20 19:50:26 +0000
a78bc60c0fMinor tweak to improve ease-of-use of iterator system.
hanna
2009-11-20 18:24:19 +0000
4fbb6d05d0Refactoring. Push the revisions to the common aligner interface down into the aligner base classes. Hack the managed implementation to support the new interface.
hanna
2009-11-20 17:08:09 +0000
d84444200bThe Unified Genotyper now sorts the sample names in the vcf that it outputs. [There was no reason to enforce that every VCF being output from the GATK should have the samples sorted, since someone might want them ordered non-alphabetically]
ebanks
2009-11-20 16:13:18 +0000
38a030f2baFinishing off data transfer conduits for single alignment generator. Misc bug fixes elsewhere.
hanna
2009-11-20 15:21:59 +0000
2a5349d886VariantAnnotator now adds dbsnp id if a dbsnp rod is supplied and it's not already set for a record
ebanks
2009-11-20 03:26:09 +0000
b434c1c240Check for null entries before adding
ebanks
2009-11-20 03:12:20 +0000
82fd824c4dContinuing improvements to unified genotyper
depristo
2009-11-20 01:39:29 +0000
33dcfc858dupdates to the paper genotyper based on Mark's comments. There's still more work to do, including more testing.
aaron
2009-11-19 23:06:49 +0000
22aaf8c5e0Added the old recalibrator integration tests to the refactored recalibrator sitting in playground.
rpoplin
2009-11-19 22:43:28 +0000
a95302fe98Single alignment generator, another checkpoint. Does generate single alignments, but some of the data still needs to plumbed through and it may leak memory.
hanna
2009-11-19 21:20:03 +0000
a972b2769fCheckpoint. Add first phase of single alignment interface.
hanna
2009-11-19 19:03:43 +0000
306f4624c6oops forgot to update the md5s
chartl
2009-11-19 18:22:29 +0000
6ba1f3321dFixed the sample mix-up bug Kiran discovered, and added a unit test in the VCF reader class (Thanks for the good example files Kiran). Also renamed the toStringRepresentation function to toStringEncoding, and added a matching method in VCFGenotypeRecord.
aaron
2009-11-19 18:17:47 +0000
b4babb82ebadding an extra bit of data to come out of CTT (number of chips with actual data)
chartl
2009-11-19 17:46:10 +0000
1fcd28bba9Update xml descriptor with proper version number.
hanna
2009-11-19 17:32:22 +0000
3f91bcf303Add staging to svn:ignore
alecw
2009-11-19 17:30:17 +0000
b3872386c9Test to ensure that ConcordanceTruthTable and those walkers which rely on it for tabulating pooled truth information from truth information of the individuals within the pool is doing that calculation correctly. Tests single het, single hom (with/without reference), together, together without reference, and a mix of everything.
chartl
2009-11-19 15:26:32 +0000
eeb3a3fffbcomments for Aaron
depristo
2009-11-19 12:56:04 +0000
7997455f38first go of the genotyper for the GATK paper. More testing and review tomorrow to call it done.
aaron
2009-11-19 07:55:24 +0000
7b957d3e2eMake the whining from Khalid's office stop already
ebanks
2009-11-19 03:04:48 +0000
85bc9d3e91(Hopefully) temporary hack: load contig information by contig name rather than contig id to avoid off-by-one errors.
hanna
2009-11-18 23:33:27 +0000
0fbd81766bCountCovariates now uses any rod of type VariationRod with the name dbsnp as the source of known variant sites to skip over. It also grabs the platform string out of the read group when deciding which algorithm to use to calculate machine cycle. In this way it can now handle multi-platform bams. I added a new covariate: PositionCovariate. This is simply the offset regardless of which platform the read came from. This will be useful for comparing between the two covariates. Finally, this message serves as a warning that I will be killing the old recalibrator tomorrow after I've updated and verified new integration tests.
rpoplin
2009-11-18 23:03:47 +0000
f667bed7fc-Don't annotate allele balance or on-off genotype if there's no genotype data -If qscore is infinity (because of precision) make a best guess instead
ebanks
2009-11-18 22:01:32 +0000
90212c643bmore effective & efficient test for SecondBaseSkew
chartl
2009-11-18 20:53:32 +0000
087e01a439minor changes for --noSLOD
ebanks
2009-11-18 18:48:01 +0000
a70cf2b763A bunch of changes needed to make outputting pooled calls possible
ebanks
2009-11-18 18:42:57 +0000
0a35c8e0ba1. The joint estimation model now constrains genotypes to be AA,AB,or BB only (i.e. to use a single alternate allele). Note that this doesn't work for the old models (point estimate or SSG) because calculations aren't divided by alternate allele. 2. Allele frequency spectrum is not emitted for single samples (since it doesn't make sense). 3. If in pooled mode, throw an exception of pool size isn't set appropriately.
ebanks
2009-11-18 17:43:15 +0000
405c6bf2c1VariantEval genotype concordance for pools! Integration test coming soon
chartl
2009-11-18 17:24:54 +0000
6fe1c337ffPileup cleanup; pooled caller v1
depristo
2009-11-18 17:03:48 +0000
f0a234ab29TableRecalibration is now much smarter about hashing calculations, taking advantage of the sequential recalibration formulation. Instead of hashing RecalDatums it hashes the empirical quality score itself. This cuts the runtime by 20 percent. TableRecalibration also now skips over reads with zero mapping quality (outputs them to the new bam but doesn't touch their base quality scores).
rpoplin
2009-11-18 16:47:44 +0000
be31d7f4ccAdded - a walker that outputs relevant information about false negatives given a bunch of hapmap individuals and corresponding integration tests for it.
chartl
2009-11-18 15:47:52 +0000
b68d6e06b7Rollback of the previous "fix" and implementation of the real fix.
chartl
2009-11-18 06:09:24 +0000
95f1be94c0Fix for the broken build:
chartl
2009-11-18 05:56:31 +0000
9fb50e9bd9Further refactoring so that pooled calling will work. Okay, Mark, you should be all set.
ebanks
2009-11-18 00:18:13 +0000
42a0bbaf46Minor reformating for pooled calling
depristo
2009-11-17 22:06:11 +0000
ec1a870905Working with byte arrays is faster than working with Strings so the Covariates now take in byte arrays. None of the Covariates themselves used the reference base so I removed it. DinucCovariate now returns a Dinuc object which implements Comparable instead of returning a String because it was too slow. CountCovariates now uses a read filter to filter out unmapped reads and allows the user to specify -cov all which will use all of the available covariates, of which there are 7 now. If no covariates are specified it defaults to ReadGroup and QualityScore, the two required covariates. Initial code in place to leave SOLID bases alone if they have bad color space quality. TableRecalibration uses @Requires to tell the GATK to not give the reference bases since they weren't being used for anything.
rpoplin
2009-11-17 21:50:52 +0000
4d9c826766Integration tests actually run on real data now. <tries to hide sheepish grin>
ebanks
2009-11-17 21:04:14 +0000
5e126875eatemporarily disable (tests are broken)
ebanks
2009-11-17 20:45:52 +0000
a048f5cdf1-Refactored JointEstimation code so that pooled calling will work -Use phred-scale for fisher strand test -Use only 2N allele frequency estimation points
ebanks
2009-11-17 20:21:15 +0000
43bd4c8e8fIgnoring deletions in the primary pileup by default was causing the primary pileup to become shorter than the secondary pileup when building up the secondary base pileup string. This fix makes sure to include the primary Ds within the pileup so that not only are the pileups guaranteed to be the same size, the same offsets will truly correspond with the same read.
chartl
2009-11-17 17:20:13 +0000
aece7fa4c7a convenience method to join a map into a single string, which I need for some VCF work. Added some documentation to the join method as well.
aaron
2009-11-17 16:50:01 +0000
21729d9311Do not print debug message when debug mode is not requested!!
asivache
2009-11-16 20:28:41 +0000
967215066dThe old CountCovariates now warns the user if they didn't supply a dbSNP rod file. Thanks Kiran for the use case.
rpoplin
2009-11-16 19:16:46 +0000
eb07c7f7f8CountCovariates now warns the user if they didn't supply a dbSNP rod file. Thanks Kiran for the use case.
rpoplin
2009-11-16 18:44:54 +0000
4558375575Stage 1 of the VariantFiltration refactoring is now complete. There now exists a parallel tool called VariantAnnotator which simply takes variant calls and annotates them with the same type of data that we used to use for filtering (e.g. DoC, allele balance). The output is a VCF with the INFO field appropriately annotated.
ebanks
2009-11-16 02:41:20 +0000
ce5034dc5dFinally reinstate the iterator-style interface. Get rid of some scaffolding code.
hanna
2009-11-16 02:34:19 +0000
103763fc84An accessor for the VCF header
kiran
2009-11-15 09:28:25 +0000
768f865035Bulletproofing code. If errors are encountering during the run, turn them into JDK exceptions whenever possible instead of random core dumps. Don't allow negative values for most of the 'bwa aln' input parameters.
hanna
2009-11-14 22:57:16 +0000
88fd762436The -rf argument is now being used for read filter and is colliding with my walkers. Changed mine to -recalFile
rpoplin
2009-11-14 19:37:46 +0000
aaron
andrewk
rpoplin
ebanks
depristo
alecw
hanna
chartl
asivache
kiran