I have previous revision to revert back to.
This is a calculator that will calculate:
P[ True base is X | read base mismatches, secondary base is Y, previous K bases are Z1,Z2,...ZK ]
where the number of pervious reference bases to take into account is user-defined. The secondary base is optional as well.
--usePreviousBases k
tells the walker to use the k previous reference bases in the transition table
--useSecondaryBase
tells the walker to use the secondary base at a locus in the transition table
these can be used together.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1816 348d0f76-0448-11de-a6fe-93d51630548a
Also a little playground walker that changes the sort order flag of a BAM file.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1805 348d0f76-0448-11de-a6fe-93d51630548a
was causing OOM issues with the new mmapping fasta file reader during large jobs.
Temporarily reverting the reader until a workaround can be found.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1801 348d0f76-0448-11de-a6fe-93d51630548a
I'll need to add some multi-sample tests in a bit...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1791 348d0f76-0448-11de-a6fe-93d51630548a
Locus Power_for_k_alleles Power_for_k-2_alleles Power_for_k-2_alleles ... Power_for_1_allele
This was a request from Jason Flannick & the T2DB group.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1786 348d0f76-0448-11de-a6fe-93d51630548a
For now, a single sample input will be special-cased in the EM model - but that will change when the EM model degenerates to the single sample output with a single sample as input. For now, the EM code for multi-samples isn't finished; I'm planning on checking that in soon.
The SingleSampleIntegrationTest now uses the UnifiedCaller instead of SSG, and so should all of you. More on that in a separate email.
Other minor cleanups added too.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1785 348d0f76-0448-11de-a6fe-93d51630548a
the loop-with-small block size. Performance improvements in loading refs are extreme;
segments can be loaded in <1ms. chr1 in its entirety can be loaded in 1.5sec (down
from 30sec).
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1781 348d0f76-0448-11de-a6fe-93d51630548a
Also, with Picard's latest changes, we need to make sure we don't double-close the sam writer.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1779 348d0f76-0448-11de-a6fe-93d51630548a
-Adding initial version of Multi-sample calculation model. This still needs much work: it needs to be cleaned up and finished. Right now, it (purposely) throws a RuntimeException after completing the EM loop.
Also:
-Fix logic in GenotypeLikelihoods.setPriors
-Add logger to the models for output
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1764 348d0f76-0448-11de-a6fe-93d51630548a
Also a fix to the GLF tests, and a correction to PrintReadsWalker to remove the close() on the output source, the source handles that itself (and you get a double close).
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1758 348d0f76-0448-11de-a6fe-93d51630548a
CalculateBaseLikelihoodsWalker.java walks through reads calculates likelihoods using SSG at each base position
CalculateAlleleLikelihoodsWalker.java walks through HLA dictionary and calculates likelihoods for allele pairs given output of CalculateBaseLikelihoodsWalker.java
CalculatePhaseLikelihoodsWalker.java walks through reads and calculates likelihoods score for allele pairs given phase information
File Readers:
BaseLikelihoodsFileReader.java reads text file of likelihoods outputted by SSG
FrequencyFileReader.java reads text file of HLA allele frequencies
PolymorphicSitesFileReader.java reads text file of polymorphic sites in the HLA dictionary
SAMFileReader.java reads a sam file (used to read HLA dictionary when in another walker)
SimilarityFileReader.java reads a text file of how similar each read is to the closest HLA allele (used to filter misaligned reads)
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1744 348d0f76-0448-11de-a6fe-93d51630548a
Some notes:
1. This design should be flexible enough to include pooled calling (for now) after discussions with Chris.
2. Using the unified caller with the SingleSampleCalculationModel emits the exact same output as SSG over all of chr20 for NA12878. Additionally, when we include the "max deletions allowed at a locus" argument (so we don't try to call SNPs at deletion sites), it removed 233 SNP calls in chr20 that were clearly indel artficts.
3. The MultiSampleEMCalculationModel is still a work in progress and will be checked in later this week.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1740 348d0f76-0448-11de-a6fe-93d51630548a
@ PooledGenotypeConcordance - changes to output, now also reports false-negatives and false-positives as interesting sites. It's been like this in my directory for ages, just never committed.
@NQSExtendedGroupsCovariantWalker - change for formatting.
@NQSTabularDistributionWalker - breaks out the full (window_size)-dimensional empirical error rate distribution by the window. So if you've got a window of size 3; the quality score sequences 22 25 23 and 22 25 24 have their own bins (each of the 40^3 sequences get one) for match and mismatch counts.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1730 348d0f76-0448-11de-a6fe-93d51630548a
Limitations:
1) Can't handle RC alignments.
2) Can't handle indels.
3) Can't handle N's in reference bases.
4) Stops at first hit.
Ran BWT over a test suite of 800k Ecoli reads. After removing alignments with indels / reads with Ns, the remaining reads were aligned with quality 'equal to' that of the alignment stored in the BAM file. In this case 'equal' quality is <= mismatches to the reference as the existing alignment stored in the BAM file.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1710 348d0f76-0448-11de-a6fe-93d51630548a
quality score of a particular base is than the quality score of its neighbors. The granularity of the binning
jumps from 32 groups to 860 groups.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1702 348d0f76-0448-11de-a6fe-93d51630548a
PARTIAL commit; new, simpler rodRefSeq will reappear in a seq.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1694 348d0f76-0448-11de-a6fe-93d51630548a
all the updated classes now make use of new SeekableRodIterator instead of RODIterator. RODIterator class deleted. This batch makes only trivial updates to tests dictated by the change in the ROD system interface. Few less trivial updates to follow. This is a partial commit; a few walkers also still need to be updated, hold on...
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1667 348d0f76-0448-11de-a6fe-93d51630548a
- Some walkers don't use the ref base, so speed up traversals by not requiring it
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1652 348d0f76-0448-11de-a6fe-93d51630548a
Also, sanity check that ref is not 'N'.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1651 348d0f76-0448-11de-a6fe-93d51630548a
The main reason for the change is that there can be (and are!) multiple RODs overlapping with a single reference base position in a single track. There can be two "trivial" RODs at the same location (e.g. samtools pileup will have two point-like records at putative indel sites: one for the reference, the other one for the indel itself). Or there can be one or more "extended" RODs (length >1), eg. dbSNP can report an indel at Z:510-525 AND a SNP at Z:515.
The ReferenceOrderedDatum object (and children) will not be changed, but it is now explicitly interpreted as a single data *record*, possibly out of many available from a given track for the current site. As long as single data record occupies one line in a data file, the new ROD system will take care of loading and keeping multiple records, including extended (length > 1) ones, and will automatically drop the records when they finally go out of scope. For one-line-per-record, multiple-records-per-site RODs, there is no need anymore for the hack used so far that involved passing ROD's own implementation of iterator through reflection mechanism (though it will still work)
* RODRecordList:
the ROD system (its iterators) will now always return a LIST of all RODs available at current position or at current query interval (see below). This class is a trivial wrapper for a list of ROD objects, with added location argument for the whole collection. The location of the RODRecordList is where the ROD system is currently sitting at: a single, current base on the reference (if next() traversal is performed), or the location of the query interval when returned by seekForward() (see below). The ROD objects themselves will have their locations set according to the original data in the file. Hence, perusing the above example of a dbSNP indel at Z:510-525 and SNP at Z:515, when moving to the position Z:515 the ROD system will return a RODRecorList with location Z:515, and with two ROD objects packaged inside, one with location Z:510-525, the other with Z:515.
*RODRecodIterator:
Almost identical to old SimpleRODIterator used by ReferenceOrderedData; this is a low-level iterator that walks over records in the data file (with a callback to ROD's ::parseLine() to parse real data)
*SeekableRODIterator:
a decorator class that wraps around Iterator<ROD> (such as RODRecordIterator) and makes the data traversable by reference position, rather than record by record. This is reimplementation of the old RODIterator. SeekableRODIterator's ::next() moves to the next position on the ref and returns all RODs overlapping with that position (as a RODRecordList). This iterator also adds a seekForward(loc) operation, that allows fast forwarding to a specified position or interval. Length > 1 query arguments (extended intervals) are fully supported by seekForward(), the returned RODRecordList wil contain all RODs overlapping with the specified interval, and the location of the returned RODRecordList object will be set to that query interval. NOTE: it is ILLEGAL to perform next() after a seekForward() query with length > 1 interval. seekForward() with point-like (length=1) interval reenables next().
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1650 348d0f76-0448-11de-a6fe-93d51630548a
- summary output now reported as a percentage rather than proportion; 2 sigfigs
- fixed minor bug where FNR was calculated over total calls rather than total variant sites
- column headers are_now_contiguous_strings
- spacing fixed
- "No Call" separated from "Ref Call" as its own column
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1644 348d0f76-0448-11de-a6fe-93d51630548a
Additional output & better output formatting. It has now undergone a good five hours of testing; and for pools of size 1 outputs exactly the same statistics as GenotypeConcordance (when GenotypeConcordance is modified to do nothing on reference='N'); and for pools of many sizes outputs close to the expected (by genetics) statistics. Looks like this is working properly.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1642 348d0f76-0448-11de-a6fe-93d51630548a
Deletion: PooledGenotypeConcordanceNew
Rewrite: PooledGenotypeConcordance. It works, and is blazing fast compared to the earlier version (1 order of magnitude speedup)! And is now entirely non-hackey, as opposed to before when there were some hacky bits.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1640 348d0f76-0448-11de-a6fe-93d51630548a
Iterates synchronously over two (name-ordered) single-end alignment SAM files with, possibly, multiple alignments per read and for each read name encountered returns pairs<all alignments for end1, all alignments for end2>
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1639 348d0f76-0448-11de-a6fe-93d51630548a
The Strand filter then needs to ignore those bases when determining bias.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1636 348d0f76-0448-11de-a6fe-93d51630548a
-added option to have "D"s inserted for deleted bases in pileup strings
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1635 348d0f76-0448-11de-a6fe-93d51630548a
@PooledGenotypeConcordanceNew - a new version of the pooled genotype concordance test for Variant Eval. Code altered to be more extensible, use a private class for handling the count tables so it doesn't gunk up the code in the test itself, and for easy debugging. The hackier methods from the original were rewritten properly. Currently computes more statistics that it outputs. Code compiles, is never called by anything, and breaks none of the tests.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1632 348d0f76-0448-11de-a6fe-93d51630548a
2. Fix ratio printouts (for params file)
3. Rename ratio filter's get counts method to avoid confusion; more changes on the way this week.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1616 348d0f76-0448-11de-a6fe-93d51630548a
don't print out intervals to be merged if they're not within the global -L intervals
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1615 348d0f76-0448-11de-a6fe-93d51630548a
@NQSMismatchCovariantWalker - Walks along the gene calculating the table
# NQS
# Q score
# mismatches at non-dbsnp sites
# total number of bases at non-dbsnp sites
And prints it out at the end.
Changes:
@PooledGenotypeConcordance now works. Takes a path to a file listing a bunch of hapmap IDs in whatever pool we want to check, reads those in, and checks for concordance by name.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1614 348d0f76-0448-11de-a6fe-93d51630548a
but mostly altering the code so it's much more readable and understandable, and much less hacky-looking.
ADDED:
@Quad: This is just like Pair, except with four fields. In the original CoverageAndPowerWalker I often used
a pair of pairs to hold things, which made the code nigh unreadable.
@SQuad: An extension of Quad for when you want to store objects of the same type. Let's you simply declare
new SQuad<X> rather than new Quad<X,X,X,X>
@ReadOffsetQuad: An extension of Quad specifically for holding two lists of reads and two lists of offsets
Supports construction from AlignmentContexts and conversion to AlignmentContexts (given
a GenomeLoc). There are methods that make it very clear what the code is doing (getSecondRead()
rather than the cryptic getThird() )
@PowerAndCoverageWalker: The new version of CoverageAndPowerWalker. If the tests all go well, then I'll remove
the old version. New to this version is the ability to give an output file directly
to the walker, so that locus information prints to the file, while the final reduce
prints to standard out. Bootstrap iterations are now a command line argument rather
than a final int; and users can instruct the walker to print out the coverage/power
statistics for both the original reads, and those reads whose quality score exceeds
a user-defined threshold.
CHANGES:
@PoolUtils: Altered methods to accept as argumetns, and return, Quad objects. Added a random partition method
for bootstrapping.
@CoverageAndPowerWalker: Altered methods to work with the new PoolUtils methods.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1602 348d0f76-0448-11de-a6fe-93d51630548a
Rollback of Variant-related changes of r1585, additional PGC code
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1586 348d0f76-0448-11de-a6fe-93d51630548a
@VariantEvalWalker - added a command line option to input a file path to a pooled call file for pooled genotype concordance checking. This string is to be passed to the PooledGenotypeConcordance object.
@AllelicVariant - added a method isPooled() to distinguish pooled AllelicVariants from unpooled ones.
@ all the rest - implemented isPooled(); for everything other than PooledEMSNProd it simply returns false, for PooledEMSNProd it returns true.
Added:
@PooledGenotypeConcordance - takes in a filepath to a pool file with the names of hapmap individuals for concordance checking with pooled calls
and does said concordance checking over all pools. Commented out as all the methods are as yet unwritten.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1585 348d0f76-0448-11de-a6fe-93d51630548a
After consulting our resident expert (Aaron), we're going to (temporarily) remove the date from the vcf output until we can come up with a better solution. However, this shouldn't cause any short-term problems because the data truly is optional.
VF test's MD5s are updated.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1580 348d0f76-0448-11de-a6fe-93d51630548a
1. VariantsToVCF can now be called statically to output VCF for a single ROD instance; this is temporary until we have a VCF ROD.
2. VariantFiltration now outputs only 2 files, both mandatory: all variants that pass filters in geli text, and all variants in VCF.
If there are any problems, go find Aaron.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1569 348d0f76-0448-11de-a6fe-93d51630548a
2) when reads with deletions are requested, adds to the pile just those: reads with 'D' over the current reference base, but not 'N'
3) next() now implements a loop: recursive forward iteration calls to next() until ref. position with non-zero coverage is encountered were OK for (short) deletions, but with long stretches of N's they end up with stack overflow
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1568 348d0f76-0448-11de-a6fe-93d51630548a
@CoverageAndPowerWalker - removed a hanging colon that was being printed after the reference position
@VariantEvalWalker - added a command line argument for pool size for eventual use in doing pooled caller evaluations. As now, the variable is unused.
@AlignmentContext - altered the scope of class variables from private to protected in order that child objects might have access to them
New Additions:
Filtered Contexts
Sometimes we want to filter or partition reads by some aspect (quality score, read direction, current base, whatever) and use only those reads as
part of the alignment context. Prior to this I've been doing the split externally and creating a new AlignmentContext object. This new approach makes
it a bit easier, as each of these objects are children of AlignmentContext, and can be instantiated from a "raw" AlignmentContext.
@FilteredAlignmentContext is an abstract class that defines the behavior. The abstract method 'filter' is called on the input AlignmentContext, filtering
those reads and offsets by whatever you can think of. The filtered reads/offsets are then maintained in the reads and offsets fields. These classes can
be passed around as AlignmentContexts themselves. Writing a new kind of read-filtered alignment context boils down to implementing the filter method.
@ReverseReadsContext - a FilteredAlignmentContext that takes only reads in the reverse direction
@ForwardReadsContext - a FilteredAlignmentContext that takes only reads in the forward direction
@QualityScoreThresholdContext - a FilteredAlignmentContext that takes only reads above a given quality score threshold (defaults to 22 if none provided).
A unit test bamfile and associated unit tests for these are in the works.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1559 348d0f76-0448-11de-a6fe-93d51630548a
-Bamboo will now run the integration tests once a day, and the normal units tests on each check-in.
-Also added a bunch of unit tests for VariantEval walker
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1555 348d0f76-0448-11de-a6fe-93d51630548a
-I also uncovered a corner case in the cleaner that for some reason was commented out but shouldn't have been. Hooray for unit tests!
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1553 348d0f76-0448-11de-a6fe-93d51630548a
Also fixed the case where MD5 sums had leading zero's clipped off
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1551 348d0f76-0448-11de-a6fe-93d51630548a
Note to self: never give project status in an svn log.
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1550 348d0f76-0448-11de-a6fe-93d51630548a
1) Moved various disparate concordance / set splitting functionalities to a new parent tool which works like VariantFiltration (i.e. people can write various modules that fit inside and can be run though it).
2) Fixed up argument parsing in VariantFiltration to use key=value format so we don't accidentally mox up values (like I had been doing).
3) Have indel rod print samples
git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1540 348d0f76-0448-11de-a6fe-93d51630548a
aaron
hanna
chartl
ebanks
jmaguire
kiran
sjia
depristo
andrewk
asivache