0da0dfa1daCosmetic change - lower-case for all command-line arguments' short names.
kiran
2010-08-08 04:12:01 +0000
eb1bb94d1cMoved the evaluation of the JEXL expressions to a point *after* the samples are subset and the INFO-field annotations are updated. I think this makes more sense than having the evaluations happen beforehand, since it seems jarring to have the JEXL expressions operate on the annotations before they're updated, and have the file contain the annotations after they're updated. Now, selecting on something like allele frequency will actually apply to the annotations that actually end up in the file, while selection on other annotations (which are carried over without modification) will act exactly the same regardless.
kiran
2010-08-08 04:09:02 +0000
594b7912f1Added a generic method for returning the complete command-line used when calling a walker, to be used in the bam/vcf headers. As requested, every possible engine/walker argument is included. I've added it to the Unified Genotyper output, so people can try it out and let me know what they think. Something that needs to be discussed in group meeting: what happens when we merge VCFs? Do we keep all of the command-lines?
ebanks
2010-08-08 03:53:07 +0000
6e389059cfAn improved version of VariantSubset and VariantSelect, meant to replace those walkers. Takes in a VCF and creates a subsetted VCF by sample(s), JEXL expressions, or both.
kiran
2010-08-08 02:57:06 +0000
41fee2d75ePublication tranches report is now the default output
depristo
2010-08-07 13:58:59 +0000
f4ffef4479Default max variants is now 5000
depristo
2010-08-07 13:58:32 +0000
80e31df40dUseful script to see the status of gsa computing resources. Crontab'd and will be arriving as email at 8 am
depristo
2010-08-07 12:36:28 +0000
b63d64bbbcBeautiful labels, better choice of dimension ranges. Supports fast loading of just first N records for testing.
depristo
2010-08-06 23:17:32 +0000
70f492a6e8Prints out trivial debugging info
depristo
2010-08-06 13:24:21 +0000
341e752c6c1) AlleleBalance is no longer a standard annotation, but the Allelic Depth (AD) is for each sample. 2) Small fixes in the VCFWriter: a) Trailing missing values weren't being removed if their count was > 1 (e.g. ".,.") b) We were handling key values that were Lists, but not Arrays. We now handle both.
ebanks
2010-08-06 12:05:14 +0000
c68625f055Fixes from Mark for the MutableContexts; this fixes the clearGenotypes() and the clearFilters() methods, and adds a method to clear the attributes. Also added is a method for creating a variant context where the attribute list is pruned to a specific subset, which can be null.
aaron
2010-08-05 22:39:51 +0000
72ae81c6deVariantContext has now moved over to Tribble, and the VCF4 parser is now the only VCF parser in town. Other changes include:
aaron
2010-08-05 18:47:53 +0000
a8d37da10bChecking in everyone's changes to the variant recalibrator. We now calculate the variant quality score as a LOD score between the true and false hypothesis. Allele Count prior is changed to be (1 - 0.5^ac). Known prior breaks out HapMap sites
rpoplin
2010-08-05 14:12:19 +0000
07addf1187Fix for Kiran: since the Variant Annotator will re-annotate on top of existing annotations it makes sense to remove old headers if they conflict with the definitions being added by VA.
ebanks
2010-08-05 06:44:39 +0000
1539791a04Fix for Kiran: when using VCFs for the comp tracks in the Annotator(s), don't put the headers from them into the output VCF.
ebanks
2010-08-05 04:45:47 +0000
227c4b10f0Bug fix for Chris: convert comp tracks to VC so that we can respect the filter field. Added an integration test to cover this.
ebanks
2010-08-05 04:13:16 +0000
84ca2f27bbBug fix for Chris: added method createPotentiallyInvalidGenomeLoc() to the GenomeLocParser that doesn't check that the contig exists in the sequence dictionary. This is crucial for lifting over from one reference to another, as sometimes contigs names change in the liftover (e.g. chrM to MT).
ebanks
2010-08-05 03:19:02 +0000
f247cbf68eI want to be the first to use the new super-cool Hidden annotation! No more telling people not to use the cleaner debugging options.
ebanks
2010-08-05 02:44:37 +0000
78bfe6ac48Added @Hidden annotation, a way to deliberately exclude experimental fields and walkers from the help system.
hanna
2010-08-05 02:26:46 +0000
82d6c5073bA simple read strand filter for potluri on get satisfaction
chartl
2010-08-04 23:23:50 +0000
d53d5ffbf6A utility class that computes running average and standard deviation for a stream of numbers it is being fed with. Updates mean/stddev on the fly and does not cache the observations, so it uses no memory and also should be stable against overflow/loss of precision. Simple unit test is also provided (does *not* stress-test the engine with millions of numbers though).
asivache
2010-08-04 21:39:02 +0000
8d8acc9faeMoving G's MyHapScore to replace the old HapScore
ebanks
2010-08-04 21:00:54 +0000
7858ffec32Spit out the error in the warning message so that Sendu can tell me what his problem is
ebanks
2010-08-04 20:40:28 +0000
5815348ebcSwitch to newer version of comp tracks (and make the trigger track a comp as well). Indel cleaning should override the interval list and only use the contig interval list; and also force jobs to go to long.
chartl
2010-08-04 20:05:27 +0000
86211b74e8Bug fix: when padding alleles in creating a Variant context from an indel, leave no-call alleles as no-call alleles.
delangel
2010-08-04 19:51:10 +0000
38e65f6e1bAdded: A VariantEval module that gives simple metrics by sample, an an abstract class that makes per-sample modules easy to write (but a little bit clunky since a class needs be defined for each data point -- see SimpleMetricsBySample as an example). AnalysisModuleScanner needed a slight update to pull in data points from parent classes for this to work (thanks Khalid for showing me how to do this). After a code review with Aaron (thanks) and ensuring integration tests pass, I am committing.
chartl
2010-08-04 19:37:39 +0000
f13d52e427Attempt to determine whether underlying filesystem supports file locking and disable on-the-fly dict and fai generation.
hanna
2010-08-04 19:28:27 +0000
1a36cb9296Can now set the maximum number of variants to see in a cluster plot (useful when you don't need to see a billion points to get an idea of what's going on. Limit applies to known and novel variants separately.
kiran
2010-08-04 18:45:24 +0000
bd27287fe7An R module that takes in a Variant Recalibration cluster file (file with '@!CLUSTER' lines in it), a tabularized VCF, and optionally a set of loci that should be examined more carefully, and emits a tremendous number of plots. For every annotation used in clustering, the distributions and pair-wise comparison (with ellipses denoting the 2-sigma cluster boundaries) are shown. Each cluster is shaded with a color proportional to its mixture coefficient.
kiran
2010-08-04 18:35:14 +0000
340bd0e2c1Removed hard-coded pointers to references
ebanks
2010-08-04 17:59:37 +0000
a47824d680A couple of type specific implementations of a single extend() method: takes an array (byte[] or short[] currently) and "extends" it to the left or to the right by the specified number of elements. Returns newly allocated array, with the content of original array copied in (if we extend by n elements to the left, then the returned array will have n default-filled elements *followed* by the content of the old array).
asivache
2010-08-04 15:30:48 +0000
012a7cf0a5mismatchCount now has a version that counts mismatches only along a part of the read (takes additional args start_on_read and length_on_read to specify the read's subsequence to be interrogated);
asivache
2010-08-04 15:27:35 +0000
e6e8a20a1e1) Fix MyHaplotypeScore to ignore 454 reads, since all those pathological non-existing indels make some sites' score blow up. If a site is only covered by 454 reads, we (hopefully) detect this graciously and just emit a score of 0.0 for the site.
delangel
2010-08-04 15:23:08 +0000
bf60ed0b25Needed it here too: warn user instead of dying if the R script cannot be executed
ebanks
2010-08-04 13:11:27 +0000
40ffe34686Warn user instead of dying if the R script cannot be executed
ebanks
2010-08-04 13:08:15 +0000
17d5e89734Now --list annotates which modules are Standard
ebanks
2010-08-03 21:00:37 +0000
2307bed742VariantEval now uses the "standard" modules only by default. You can add other modules with the -E argument and not use all of the standard ones with -noStandard (they can be added back individually with -E).
ebanks
2010-08-03 16:51:10 +0000
a7ff9caf54Added sanity check against bad people and/or crazy big indels at edges of ref context
ebanks
2010-08-03 05:37:17 +0000
5f1b67c1deCoping out and forcing the entire GATK (and associated JVM) to use US English locale. Method to force JVM into proper locale exists in CommandLineProgram and is disabled by default, but implementers of CommandLineProgram can opt in to the forced US locale by calling a static method.
hanna
2010-08-03 03:48:26 +0000
b5b2c19124Updated resources package descriptor with dbsnp 129 for b37.
hanna
2010-08-03 02:42:48 +0000
c203e0fb02Added JEXL support for hetCount, homRefCount, and homVarCount in VCs.
depristo
2010-08-02 12:24:11 +0000
7fab5c0a8fsupport for -singleton_fp_rate arguments to variant recalibrator instead of the pop.gen. AF prior. Worth experimenting with Ryan.
depristo
2010-07-31 21:17:47 +0000
6dcb63888dBe smart about the headers.
chartl
2010-07-30 16:11:16 +0000
f20cdbe60aModified to work with MT containing VCFs.
chartl
2010-07-30 14:59:59 +0000
6d91cd587eBe explicitly clear about which options are for debugging purposes only and shouldn't be used if your username is not ebanks@broad. If only we had a @hidden annotation option for args...
ebanks
2010-07-30 14:18:31 +0000
ac8048f17bSupport for automated selects for tranches in variant eval -- use -tf to make tranch-specific ve outputs. ApplyVariantCuts with tranche reading functions for general use, along with todo for ryan. CombineVariants now has --filteredAreUncalled and will treat filtered snps in input VCFs are uncalled, and so won't emit -filteredInOther set features
depristo
2010-07-30 14:16:43 +0000
9231d13252Minor modification: adding an argument to make slightly more general.
chartl
2010-07-30 05:20:20 +0000
db54d63fc7Hahaha yes, ownage. This now works.
chartl
2010-07-30 04:23:02 +0000
3e3f8c7692Simple count intervals walker, as per my recent email to GSAMembers. Never use this. It doesn't behave the way you think it does.
chartl
2010-07-30 03:39:23 +0000
9132c98eecSlightly smarter interval list dealing (whole exome intervals are .interval_list, whole genome are .interval.list). Also use BTI with the Genomic Annotator
chartl
2010-07-29 22:04:02 +0000
54d93f63d2Hacky fix for LSF confusion -- submitted jobs check to see if their directory exists, despite depending on the job which creates said directory. Filter strings now have escaped quotes.
chartl
2010-07-29 21:35:50 +0000
0f9baa2e94Ha ha ha ha ha
chartl
2010-07-29 20:48:35 +0000
ba1a330293Corrected location and made more explicit the error message thrown if someone tries to read a VCF 3.3 file with indels, which is not supported.
delangel
2010-07-29 20:02:47 +0000
735ef19dc8Added option to sleep after creating temporary directories.
kshakir
2010-07-29 19:53:17 +0000
82c37fceb5Create intermediate directories and don't error if the directory already exists.
kshakir
2010-07-29 19:45:12 +0000
7a5ee485d2Full pipeline now works through DAG creation. First draft; more work to do to make it cleaner and better command-line input handling (and properties handling); but the DAG is rendered and looks good.
chartl
2010-07-29 19:36:17 +0000
5af986e0c1Add an integration test for Beagle (one for ProduceBeagleInput and one for BeagleOutputToVCFWalker)
delangel
2010-07-29 18:49:22 +0000
4d4cf6e1dcUpdates to calling pipeline
chartl
2010-07-29 18:37:20 +0000
52f24c86faScript to split a provided interval list into contigs. Excesses will be dropped into the last provided file. Works like splitIntervals.sh. This is for Queue.
chartl
2010-07-29 16:42:53 +0000
f35e6d73b4Actually name the class the name of the file. (Clearly created by cp)
chartl
2010-07-28 15:18:04 +0000
cd9395fa14Since Picard's FixMateInformation merges, fixes mates, and sorts, allow it to be used as a gather function.
chartl
2010-07-28 15:09:19 +0000
e1a34685fdAdd back MyHaplotypeScore as a new implementation for HaplotypeScore, this time as a non-standard annotation. Implementaiton is also better, it computes better consensus haplotypes, ranks them by sum of quality score.
delangel
2010-07-27 21:23:19 +0000
6c93b13428A Java sizeof, implemented using the Java instrumentation API. Can either get the memory consumed either only by a single object or by a single object and all the references it contains. Requires a command-line change to add a Java agent to the command-line; see the Sizeof.java javadoc for details.
hanna
2010-07-27 18:44:15 +0000
f5566a6593Knocking out some quick findBugs.
rpoplin
2010-07-27 14:10:59 +0000
894623858dOK, bad idea to add new temporary annotation - revert to keep integration tests hapy.
delangel
2010-07-27 12:07:13 +0000
71bfb1ee35First redesign of HaplotypeScore - now, a different approach is taken to build possible haplotypes at a site: first, all possible haplotypes consistent with reads are formed (reference is not used). After this list has been formed, it is ranked according to the number of reads that are consistent with it and the two most popular haplotypes are chosen. this reduces to the old method in typical cases, but it builds haplotypes correctly if there are two variants close by within a context window.
delangel
2010-07-27 10:54:56 +0000
cffebcc867Small utility walker used for production of the Beagle data processing paper section. Walker will print out to output file, for every site common to a reference vcf and an eval vcf, a given sample's depth, hapmap AC and AF and pre/post Beagle genotype as well as corresponding reference (e.g. Hapmap) genotype.
delangel
2010-07-27 03:00:17 +0000
1d9ed1e214Cleanup of old VCFRecord code
ebanks
2010-07-27 02:56:47 +0000
9667942e52fix for Ryan's issue: we also need to sync when we store a resource.
aaron
2010-07-26 22:17:47 +0000
8b072b59e2Returning index dumping functionality in BAMFileStat to a useable state.
hanna
2010-07-26 20:03:50 +0000
19ad44d332Minor improvements to CombineVariants to handle the complex case from Chris. IntegrationTest of complex case.
depristo
2010-07-25 13:46:11 +0000
56de475f11Based on feedback from non-GSA users, who claim that our exceptions are 'scary and overwhelming,' I've cleaned up the error message to first describe the error and what users should do and then ask them to copy the subsequent stack trace into their GetSatisfaction posting.
ebanks
2010-07-25 03:57:44 +0000
9bd8a2685bBecause the performance tests were busted on LSF, no one caught this error until now: when Matt changed over the contract for the AlignmentContext, this line needed to get updated too. All is well now.
ebanks
2010-07-25 02:53:01 +0000
b551eaf8fdActually commit the code that makes variant eval run in a reasonable amount of time.
depristo
2010-07-24 17:32:03 +0000
b0b37c3476No handles (I believe) reference only VCs correctly
depristo
2010-07-23 23:09:23 +0000
e21376219dUpdates to CombineVariants for Tim. -setKey can be null. Integrationtests for -setKey foo and -setKey null.
depristo
2010-07-23 22:35:52 +0000
26bb1cd9ceFix broken test correctly
delangel
2010-07-23 20:47:41 +0000
4fc1db7aafChange interface to VCFWriter add() method to take only 1 byte from reference (since that's the only thing it needs), to prevent bugs like having people call it with ref.addBases() which is wrong (since it provides bases starting from the left of reference context window).
delangel
2010-07-23 20:24:03 +0000
b3fd145161fix for a bug deep in the tribble indexing: if you had a single record in the first contig, the second contig's index blocks would point to the wrong file seek location, and you'd see no features in that contig. Thanks to Mark for finding this. I'm not rev'ing the index version (which would cause all indexes to be rebuilt), since this seems like a pretty rare edge case.
aaron
2010-07-23 18:39:55 +0000
33090629eaVariantEval can now see the EvaluationContext group objects, so they can decide if/when to print interesting sites. GenotypeConcordance has a hard-coded option to print FNs that is on the way to being generally useful. VCFWriter now uses the US locale for formatting floating point numbers; I believe this fixes a long-standing annoyance. Italian guys will check on this.
depristo
2010-07-23 17:16:50 +0000
77cace3affIt's probably a good idea to look carefully at what you've done before actually committing
ebanks
2010-07-23 17:06:54 +0000
44f9a631d6Adding CombineVariants to release. Pining for the days when all of core/playground will be part of release by default...
ebanks
2010-07-23 16:53:18 +0000
5eef15cfdfa) Bad bug fix to CombineVariants: when indels were being merged, the reference base provided was wrong - ref.getBases()[0] was being used, but this returns bease at start of window. Instead, the reference at current locus should be used.
delangel
2010-07-23 15:13:47 +0000
4ff8b8fc0e1. Fixing a bug that Mark found where indel-containing clipped reads would get an original cigar tag even when they didn't actually get modified. 2. Added some useful logging messages. 3. Added a oneoffs walker to calculate the number of realigned reads and intervals containing them.
ebanks
2010-07-23 14:24:01 +0000
973934f769Depth of coverage now uses longs rather than ints. We can now successfully run on the Lepidosiren paradoxa genome. (about 80 GB)
chartl
2010-07-23 14:14:12 +0000
536399eaa0Improvements to variant combine. Now calculates AC/AN/AF correctly by calling into the VariantAnnotator engine. Automatically removes annotations that are inconsistent across incoming VCs (in simpleMerge). TODO bug fix for Guillermo/Eric.
depristo
2010-07-23 13:33:11 +0000
9579aace1fupdates to code dependent on Tribble, as well as the following Tribble changes:
aaron
2010-07-23 01:56:10 +0000
kiran
ebanks
depristo
aaron
fromer
rpoplin
hanna
chartl
asivache
delangel
kshakir