v3.2.2

The correct files to download are abismal-3.2.2.tar.gz or abismal-3.2.2.zip
Please do not download the "Source code" files. Those are generated by GitHub but lack certain files needed to build abismal.

What's Changed

• Users can specify a set of target regions that are indexed; the rest of the genome is excluded.
• Allows for a gzip compressed genome file when indexing.
• Build system updated to facilitate building binaries for macOS through the Conda.

Full Changelog: v3.2.1...v3.2.2

v3.2.1

The correct files to download are either abismal-3.2.1.zip or abismal-3.2.1.tar.gz
Please do not download the "Source code" files. They are automatically generated by GitHub and incomplete.

What's Changed

• Minor bug fix with progress bar for abismal. Otherwise this version functions the same as v3.2.0.

v3.2.0

The correct files to download are either abismal-3.2.0.zip or abismal-3.2.0.tar.gz
Please do not download the "Source code" files. They are automatically generated by GitHub and incomplete.

What's Changed

• Minor optimizations in local alignment by @andrewdavidsmith in #21
• Added regression tests to run with make check by @andrewdavidsmith in #23
• Improved sensitivity for very low quality paired-end reads, costing up to 3.9% total speed, by @andrewdavidsmith in #25
• Direct BAM format output by @andrewdavidsmith in #27

New Contributors

• @masarunakajima made their first contribution in #20
• @andrewdavidsmith made their first contribution in #21

Full Changelog: v3.1.1...v3.2.0

v3.1.1

The correct files to download are either abismal-3.1.1.zip or abismal-3.1.1.tar.gz
Please do not download the "Source code" files automatically generated by github, as they do not contain all files necessary for compilation

This is a patch release, as the prior release (v3.1.0) did not properly enable multithreading with some OS/compiler combinations (e.g., macOS with the clang obtained through Xcode).

v3.1.0

The correct files to download are either abismal-3.1.0.zip or abismal-3.1.0.tar.gz
Please do not download the "Source code" files automatically generated by github, as they do not contain all files necessary for compilation

This release contains minor optimizations and bug fixes as described below

• abismal was behaving unexpectedly if, by mistake, paired-end read files did not contain the same number of reads. Now it will report an error informing the user should that be the case.
• fixed a couple of cases where reads could be reporting in inconsistent CIGAR lengths.
• if run using verbose (-v), abismal reports the number of threads used, and warns the user if the number of threads requested is more than the number of cores available in the machine.
• some minor optimizations to the banded Smith-Waterman algorithm should make this version run about 5% faster than 3.0.0

v3.0.0

The correct files to download are either abismal-3.0.0.zip or abismal-3.0.0.tar.gz
Please do not download the "Source code" files automatically generated by github, as they do not contain all files necessary for compilation

This version of abismal implements new indexing and encoding strategies that will be detailed in a future preprint. It results in faster and more accurate mapping times than 3.0.0.

Please note that your genome indices must be rebuilt to use this new version. Indices built on 2.0.0 or 1.0.0 no longer work

Indexing should take at most 20 mins on a single-core machine for the human genome.

This new release also resolves the following issues:

#5 : abismal now imposes a max read length of 65535 and stops mapping if the user attempts to map a read that is too long.
#8 : we no longer use parallel/algorithm for parallel sorting, so it should be easier to install abismal on OSX machines.

v2.0.0

The correct files to download are either abismal-2.0.0.zip or abismal-2.0.0.tar.gz
Please do not download the "Source code" files automatically generated by github, as they do not contain all files necessary for compilation

This version of abismal has several improvements over abismal-1.0.0 (described below). Please note that any index built with 1.0.0 must be rebuilt, as they are not compatible with the current release.

List of changes:

• The new index uses less memory. For hg38, the memory footprint for mapping is reduced from 3.5 GB to 2.7 GB
• Abismal 2.0.0 now performs full alignments
• Abismal 2.0.0 has algorithmic guarantees on exact matches. If a read matches the reference genome exactly somewhere, the match is guaranteed to be found.
• Abismal 2.0.0 is faster in smaller genomes. Search parameters used internally are based on genome k-mer frequencies rather than "hard-coded", which improves speed in genomes like arabidopsis, chicken and bacteria.
• A documentation is now available describing the SAM output format, a detailed description of each parameter and several examples of use.

Some issues are also resolved relative to the previous version:

• 9c93354 : resolves an issue where abismal crashes if a long read is passed to it. If a read is larger than 65535, abismal will report an error message
• the percent_unmapped report now shows 100 minus the percent_mapped (whereas before it showed 100 - percent_unique)
• the error_rate is no longer a percentage, but rather a fraction

v1.0.0

The correct files to download are either abismal-1.0.0.zip or abismal-1.0.0.tar.gz
Please do not download the "Source code" files automatically generated by github, as they do not contain all files necessary for compilation

This version of abismal contains several bug fixes addressed since the release of version 0.3.0, as well as some added functionality. Most significant changes are listed below:

• 50576cd : If users build abismal with autotools, they will have more compile information to report in case of bugs
• 2b3cb7e : abismal was crashing if a malformatted FASTQ input contained an empty read name. This bug was fixed
• 3722a01 : When indexing a reference genome with verbose on, users will be informed of how much RAM is necessary to map reads to the indexed reference
• 25e7288 : We fixed the bug on issue #4 , where the number of unmapped reads was being reported as total - unique, when it should be total - mapped. In this commit we also reorganized the mapping statistics in a more intuitive way, where percentages and numbers are grouped together for easier comparison

v0.3.0

Version 0.3.0 provides some bug fixes and algorithmic improvements, but still operates under the same algorithmic structure as 0.2.0. The major changes are

• This version includes some modifications to the algorithm geared toward including accuracy. We also removed some parameters like -c, which are now automatically inferred from the genome. IMPORTANT: Indices must be re-built with this version of abismal. Older indices are no longer compatible.
• Added a -g flag where you can pass the genome directly, so that you don't need a separate indexing step. The program abismalidx is still available, and we recommend indexing reference genomes that will be used multiple times.
• More index statistics are provided during indexing in verbose mode, where the distribution of two-letter k-mers shed light on how heterogeneous the genome is.

Please do download the files named either abismal-0.3.0.tar.gz or abismal-0.3.0.zip. Please do not download the "Source code" files auto-generated by github as they do not contain the necessary files to compile the program.

v0.2.0

This release has major algorithmic improvements to abismal that decrease memory usage and improve accuracy, namely:

(1) minimizers are used as a method to reduce memory usage and disk space in indexing. The shifts parameter no longer exists, as it is fully defined by the minimizer window

(2) more alignments are made in both single- and paired-end mode. In single end, up to 100 valid hits (with >40% edit distance) are aligned. Alignments are not performed if exact matches are found

(3) best candidates are not selected based on edit distance, but on alignment score. Furthermore, if an alignment soft-clip bases, we use the soft-clipped read length as criterion to decide if the read is valid (i.e. passes the minimum threshold to be reported)