allow setting max depth

previously regions with > 1000 depth were filtered out. That can now be set with an environment variable:

export SMOOVE_MAX_DEPTH=20000
smoove call ...

larger chroms, more samples

as of this release, the argument to smoove paste can be a file ending with .list that has one file per sample.

v0.2.7

• use csi index for bams to support larger organisms
• fix ordering of error-checking (#146)
• support for longer file-lists (#151)

bugfix

v0.2.6

• fix name collisions with cram (#134)

shorten run-time for problematic samples

v0.2.5

• slight sensitivity increase
• fix bug in VCF parser that could manifest inside of smoove annotate
• reduce run time for problematic samples. using a kdtree.
• @brwnj fixed the docker container so that conversion to singularity works without futzing.

general bug-fixes

v0.2.4

• add ##reference=$fasta to vcf header. (#58)
• fix smoove annotate for recent ensembl GFFs which lack transcript features
• fix bug in lumpy_filter for some versions of GCC (arq5x/lumpy-sv#300)
• just report message on weird alignments instead of failing (#68)
• lower lumpy z cutoff from 4 to 2.75. because of distribution of fragment lengths, this does
  not affect much, but does improve sensitivity after duphold-DHFFC filtering slightly.

diagnostics and lower FPR

v0.2.3

• save disc and split counts and output a plot as QC
• annotate: annotate with up and downstream 5KB. so any gene within 5kb will be annotated.
• smoove merge now outputs a plot that shows the number of split and discordant reads for each sample.
• smoove now filters about 25% more split reads by looking at how the read is consumed
  this results in a 5% reduction in false positive genome-in-a-bottle deletions > 1KB while leaving
  the true positives unchanged relative to version 0.2.2. (Thanks @ernfrid for feedback on this feature).

v0.2.2

NOTE that this release requires a very recent version of mosdepth

v0.2.2

• huge reduction in number of discordant reads sent to lumpy for some problematic samples.
  this is work by @ernfrid which reduces lumpy runtime for all cases and dramatically lowers
  it for cases that were previously problematic due to large numbers of discordant reads.
  On our Genome in a Bottle test sample, this drops the number of discordant reads
  from 1,434,810 to 332,384; a 4.3 fold reduction while leaving the final smoove output unchanged
  except for 2 BNDs that were removed. This reduction in reads will be even more dramatic
  in problematic samples.
• switch to use svtyper>=0.7.0 with --max_ci_dist parameter. To avoid this set:
  export SMOOVE_NO_MAX_CI=xx # any value will work; however, it's higly recommended
  to use as-is.
• cnvnator changes and filtering. (still requires github.com/brentp/CNVnator fork).
• bugfix for case when no output directory was specified.
• use mosdepth with --fast-mode NOTE this requires the latest version of mosdepth (0.2.4 or greater)

general improvements in sensitivity for small events

this release was driven by feedback from Dave Larson (@ernfrid) and Ira Hall's group. It results in improved sensitivity for smaller events (at a small cost of specificity). It also fixes a bug with the duphold runner.

v0.2.1

• fix bug in smoove duphold for samples > threads that resulted in stalling
• smoove is now more discerning about reads that are soft-clipped on both ends as
  these could be due to inversions. if a read is not flipped relative to its mate
  and it is soft-clipped on both ends, it is still removed.
• in some cases reads with a large NM (mismatches) were not filtered because they
  had an unexpected type (uint8). smoove now checks more types and for some bams
  will remove more records which improves specificity.
• smoove will now be more conservative with NM counting. an insertion of 7 bases
  is counted as an NM of 7. smoove will now correct this to an NM of 1 so it counts
  the number of events rather than the number of bases of each event.
• simplify rules for filtering and make them less strict.

v0.2.0

v0.2.0

• add duphold annotations. (requires duphold v0.0.5 or higher)

v0.1.11

• flip start and end for small percent of cases where that's a problem
• expose min sample weight to allow adjusting required support for each variant

s/Number/Integer/

the type of SHQ and MSHQ was invalid in previous versions that is fixed in this release.