v1.5.r0

• Caching of discordant read pairs during SV identification to reduce lookups on bam file. This gives a speedup of about 2x on average, but likely better returns on samples with incredibly large numbers of SVs.
• Optimizations to fetching of downsampled reads (alters results slightly as randomness of downsampled reads is updated)
• Optimization to intersection detection implemented in AA. Results in a small speedup.
• Add support for parsing of SV VCF generated in GRIDSS having extra column for normal.
• Logging improvements
• Typo fixes (PR by cmdcolin)

v1.4.r2

• provides better handling of convergence failures during Mosek optimization
• fix issue with not logging cycle decomposition info at the right place

v1.4.r1

v1.4.r1 adds the following updates:

• Reduces inverted sequence artifacts by about 50%, providing quality improvements to heavily artifacted samples.
• Bugfix to counting of foldback support reads & boundaries of foldback SVs.
• Adds --foldback_pair_support_min argument to allow users to adjust read support requirement for foldback SV discovery.
• Provides more unified behavior between python2 & 3 usage when computing read pair support requirements.
• Improves numerical stability of meanshift p-value calculation.

v1.3.r8

AA 1.3.r8 adds breakpoint microhomology detection from alignments marked as supplementary (not just secondary). Also tweaks sashimi plot visualization of CN.

v1.3.r7

• Refine granularity of amplicon plot CN segmentation for intervals between 50 to 1000 kbp. Changes way the cnseg files are named (see README). Prevents dramatic undersegmentation in some cases.
• Slightly relaxes mosek convergence criteria to prevent mosek errors reported in rare cases (thanks Michal Adamazek @ MOSEK)
• Also allow "." in FILTER field of external SV VCF file for passing calls (not just "PASS")

v1.3.r6

• --sv_vcf argument which allows users to augment AA's SV detection with their own SV calls provided in a VCF format. This does not remove SVs discovered by AA. It only adds those from the VCF not already found by AA.
• --sv_vcf_no_filter to permit calls from --sv_vcf even if they are not set to 'PASS' or '.' .
• Automated protection against improperly-formatted inputs
• Reduces bugs created when AA is rerun into the same directory with existing files having the same sample name but launched with different input files.
• Bugfix for edge case where AA does not properly expand a newly discovered interval if a discovered SV lands exactly on the endpoint of the explored interval.
• Refactor Readme and standalone install info
• Provide explicit LICENSE file

v1.3.r5

Version 1.3.r5

Provides better compatibility with the AmpliconSuite-pipeline Singularity image and versions of Mosek installed via pip/conda.

v1.3.r4

• Fixes a longstanding bug in the visualizations, where the coverage values shown in the plot are wildly incorrect (too high). This issue did not affect the graph/cycles files, just the plots. See diagram below.
• Smoother coverage plotting (less jaggedness in visualization for smaller segments).
• Adds a test at the beginning to ensure Mosek is correctly installed and the license is in place. Previously it would fail silently partway through the run, making it appear similar to a no-amp sample.
• Improve handling of the coverage.stats file.

The other outputs are the same, however there is a modest speedup to AA following these improvements. Plotting amplicons is ~4x faster. For moderately complex AA samples, it seems the total speedup is around 10%-15%, and for simple cases the total speedup is around 2x.

v1.3.r3

Version 1.3.r3.

v1.3.r2

Version 1.3.r2 sets a deterministic seed for read downsampling, which can be disabled by setting --random_seed.