v2.04

• Semi global aligner speed optimization

v2.03

• Require assembled contigs to be "soft anchored" near contig ends against the reference.
• Filter low mapq reads from assembly and alt contig generation appropriately.
• Various improvements to Cadabra.
• Introduced simple splice junction counter.

v2.02

• Addressed bug where some processing regions were being dropped
• Skip processing of intronic regions for RNA
• Various fixes to indel left alignment - this was causing several good realignments to be dropped since the switchover to semi-global contig alignment due to ambiguous cases.
• Set mapq of originally unmapped reads when realigned to max mapq (allow for inclusion in variant calling)
• Prioritize assembled contigs over secondary contigs when realignment is ambiguous
• Semi-global backtrack optimization

v2.01

• First pass at replacement of Smith Waterman alignment with semi-global Needleman-Wunsch (param names to be changed later)
• Left align indels in contigs (contigs containing multiple indels need additional work)
• Filter contigs with indels near ends
• Optionally identify consensus sequence from soft clipped reads piled up at same position (--cons)
• Default to read mismatch rate max of 5% instead of 5 total mismatches
• Parameterized max mismatch rate (--mmr) and window size / window overlap​ (--ws)

v2.00

• Upgraded to current htsjdk
• Use Intel GKL for compression
• Fixed bug where a read overlapping multiple regions was only identified as overlapping only one
• Randomize tied contig scores when using soft clipped reads as contigs
• Added --obs for parameterization of observed indel usage (was previously driven by --sws)
• Allow out of region reads to be remapped to nearby regions
• Sort output alignments
• Parameterized tmpdir location (this can now grow large)

v0.09

• Use more reasonable defaults for soft clip contig generation
• Exposed soft clip contig generation params

v0.08

Remove syserr statement. Upped to 0.08

v0.07

• Allow non-assembled contigs to be shorter than max read length.

v0.05

• Stability improvements including addressing frequent OOM errors caused by too many junction permutations.
• Don't drop reads when max cached reads threshold reached
• Increase max cached reads threshold to 500,000

v0.03

Cap max reads in memory per sample / thread.