2.3.0

• The RResolver algorithm is now part of the default pipeline. RResolver resolves repeats using short reads and generates assemblies with better contiguity and often fewer misassemblies.

rresolver-release

• Added the RResolver component to the ABySS pipeline. This release updates the 2.2.4 ABySS version and is used as a reference for the RResolver manuscript.

2.2.5

Release version 2.2.5

General:

• Resolve various compilation errors in newer versions of clang
• Use ntHash's 0th hash as the default hash instead of the 1st hash
• Added optional RResolver module, not currently part of the ABySS assembly pipeline

abyss-rresolver-short:

• Improves genome assemblies at unitig stage by using a sliding window at read size level
  across repeats to determine which paths are correct
• For further information: consult http://www.birollab.ca/assets/posts/195_Nikolic_Vladimir_HiTSeq_ISMB2020.pdf

misc:

• Extract all tags in a SAM file

2.2.4

• Release version 2.2.4

General:

• Refactor deprecated functions in clang-8

Sealer:

• Remove unsupported -D option from help page

abyss-bloom:

• Add counting Bloom Filter instruction to help page

abyss-bloom-dbg:

• Report coverage information of unitigs

2.2.3

• Release version 2.2.3
• Revert memory consumption of Bloom filters to pre 2.2.0 behaviour.
  ABySS will now share the specified memory among all Bloom filters
  instead of just the counting Bloom filter.
• Fix gcc-9 compilation warnings

2.2.2

• Release version 2.2.2
• Fix abyss-overlap for 32-bit systems

2.2.1

• Release version 2.2.1
• Fix abyss-bloom for macOS

2.2.0

• Release version 2.2.0
• Construct a counting bloom filter instead of a cascading bloom filter.

abyss-bloom:

• Add 'counting' as valid argument to '-t' option to build a counting
  bloom filter

2.1.5

• Release version 2.1.5
• Compiler fixes and increase stack size limits to avoid stack overflows.

abyss-pe:

• Add 'ulimit' statements to the Makefile to increase a thread's stack size to 64MB.

2.1.4

This release provides major improvements to Bloom filter assembly contiguity
and correctness. Bloom filter assemblies now have equivalent scaffold contiguity
and better correctness than MPI assemblies of the same data, while still
requiring less than 1/10th of the memory. On human, Bloom filter assembly times
are still a few hours longer than MPI assemblies (e.g. 17 hours vs. 13 hours,
using 48 threads).

abyss-pe:

• Change default value of m from 50 => 0, which has the effect
  of disallowing sequence overlaps < k-1 bp. QUAST tests on E. coli /
  C. elegans / H. sapiens showed that both contiguity and
  correctness were improved by allowing only overlaps of k-1 bp
  between sequence ends.