We are pleased to announce the release of version 2.4 of the SeqAn library and updated versions of the official SeqAn-based applications.
The alignment module has received significant updates and a parallised interface. Beyond simple multi-threading the alignment module can utilise the vector extensions of modern CPUs (SSE4, AVX2, AVX512) resulting in very impressive speed-ups, especially on long sequences. In fact speed-ups of ~2000x over the serial case have been measured (AVX512 and 40 threads VS no vectorisation and one thread). The details are due to be published soon.
The indexing module now also offers a parallelised interface which distributes multiple searches over the available CPU cores. Furthermore “optimal search schemes” can now be used for performing even faster searches with multiple errors. A preprint documenting these algorithms is available here:
The ReducedAminoAcid module now offers more reductions based on more recent research and VCF I/O is now compliant to version 4.2 of the specification, i.e. rows with only eight fields are accepted.
Major work went into fixing SeqAn platform support on non-canonical and big endian platforms, we are proud that also binary formats like BAM pass all tests on platforms like PowerPC64 and ARM64. All main release architectures of Debian GNU/Linux are now well supported and automatically tested. Sparc64 is one of the few platforms known not to work.
We have improved compatibility with new compilers and the C++17 standard so SeqAn should generate no warnings in 2018, even with stricter warning levels.
Finally we would like to announce that is going to be the last feature release of SeqAn2. Most of our effort is going into the development of SeqAn3, now, another announcement will cover this shortly. Patch-level releases of SeqAn2 may, however, be released if we find serious bugs or regressions.