v2.14

• Avoid crash when known variant VCF not specified for RNA.
• Limit number of possible junctions in a region.
• Skip extremely high depth sites when calling.

v2.13

• Improved handling of known indel sites in RNA
• Minor SAM spec corrections for STAR bams
• Support reading observed junctions directly from bam input
• Do not attempt to use reads with length of zero (from Lenbok)
• Added --gc option. Skips remapping if eligible regions do not contain at least one contig with an indel or splice junction (experimental)
• Added option to not remap reads with new cigar containing deletion bracketed by introns (experimental)

v2.12

• Cadabra - Close SamReader after processing chromosome (avoid too many open files errors)
• Added guard around indel component size of 0 for complex indels
• Ignore track and browser in bed file (from dr-artio)

v2.11

• Decreased memory footprint during final sort / mate fix stage
• Stop dropping reads realigned across chunk boundaries
• Assign mate info properly when one half of fragment was originally unmapped
• Adhere to VCF spec when calling variants

v2.10

• Correct issues related to single end processing
• Avoid NegativeArraySizeException when small target regions specified

Known issues:

• Occasional reads being dropped / mate not updated not yet resolved.

v2.09

Note: Previous releases were compiled on Centos 6. This release is compiled on Centos 7.

• Don't allow read to map beyond the end of a chromosome
• Default GKL to off. Appears to cause stability issues in some cases. (Re-enable using --gkl)
• Implemented disk backed sort / mate fix. Should address out of memory errors during sort phase.
• Added additional logging around read buffer problems causing potential crashes
• Update edit distance using STAR's nM tag when appropriate.
• Improved handling of exon skipping junctions
• Cadabra: Improved handling of reference bias when calling at repeat locations

Known issues:

• Still seeing a couple of reads with mate not updated properly in WGS test.
• A small number of reads are being dropped in RNA testing.

v2.08

Speed and accuracy are generally improved in this release (particularly for RNA).

• Merge and error correct overlapping reads for use in assembly / contig generation.
• Improvements to assembly triggers. Assembly is much less frequent now.
• Avoid adapter read through triggering uncessary assemblies.
• More sensitive assembly in place (less aggressive pruning / downsampling)
• Improved handling of unannotated exon skipping junctions.
• Default --dist to 1000 for DNA (Improves RAM usage when sorting large files). Recommend using 500000 for RNA to allow movement across large introns.
• Handle STAR's nM tag
• Improved options for Cadabra
• Skip decoy chromosomes / unplaced contigs, etc. by default.
• Identify smallest repeat unit when calculating repeat period
• RNA speed optimization : Limit # contigs in the face of large number of junction permutations
• Improvement to ISPAN annotation for inserts
• Fixed bug where RNA regions where being skipped due to large splice junctions
• Added option to downgrade mapping quality when a read maps equally well to reference and a contig (potentially useful for downstream callers that handle STRs / repeats naively).
• Treat deletion adjacent to intron as a potential alternative intron.
• Avoid religning very noisy reads

v2.07

• Allow remap of unclipped portion of soft clipped reads when entire read does not realign
• Update read mate info properly (output BAM files should now pass Picard's ValidateSamFile )
• Align non-assembled contigs to all junction combos
• Default mapq filter to 20
• Don't allow contig to exceed max buffer size (was causing occasional segmentation faults)
• Added --nosort option. Use to disable final sorting (and mate fixing) step.
• Added --dist option. Use to shrink max distance a read can be moved (also impacts sorting)

v2.06

• Now using more conservative assembly triggers
• Added trigger for more sensitive assembly (enables modest improvement for some longer inserts)
• Accomodate reads spanning multiple indels for observed indel contig generation
• Correction to affine gap penalty calculations in contig alignment
• Contigs now left aligned during semi-global alignment instead of post-hoc
• Ported semi-global aligner to C (for speed improvement)
• Parameterized final BAM compression level (consider using 1 for intermediate files that are to be deleted)
• Parameterize anchor length / mismatches (consider using this along with --cons for amplicon data)
• Parameterize max reads per region

v2.05

• Enable usage of high quality soft clipping and observed indels to generate putative contigs by default
  ** Note: parameters around this behavior have changed

• Remap reads equal to min mapq

• Output unmapped read pairs in final BAM

• Added deBruijn graph simplification

• Cadabra - only use primary alignments for calling

• Various speed optimizations
  ** More conservative assembly triggers
  ** Optimization of read to contig mapping
  ** sparse_hash --> dense_hash
  ** Introduce smaller chromosome "chunks" to improve parallelization
  ** Parallelize final BAM sorting/writing in multi-BAM cases
  ** More selective permuting of regional splice junctions combinations
  ** Discard low scoring contigs during assembly
  ** Cadabra - added multi-threading