DIAMOND v2.2.3

• Clustering workflows now by default apply BLAST-like compositional matrix adjust, unconditionally, instead of compositional score correction described by Hauser et al., 2016.
• Clustering workflows now by default no longer apply tantan soft masking but conservative SEG masking corresponding to BLAST defaults.
• Added the --comp-based-stats 6 option to enable compositional matrix adjust using an optimized Sinkhorn–Knopp based algorithm.
• Added the --relative-entropy-tolerance option to control the convergence tolerance of the Sinkhorn–Knopp algorithm when using --comp-based-stats 6.
• Added the option --taxdump to set search path for NCBI taxdump files other than the database directory.
• Fixed a non-determinism in the greedy vertex cover workflow and transitively in the clustering workflows (not affecting correctness of results).
• Fixed a non-determinism in the clustering workflows (not affecting correctness of results).
• Added the option --masking seg-all to apply SEG masking to both query and target sequences.
• Fixed --comp-based-stats 2-5 not correctly applying SEG masking of targets by default instead of tantan masking.
• Added the option --reps to set output FASTA file for representative sequences for the clustering workflows (contributed by @bbuschkaemper).
• Fixed a crash in the clustering workflows for large input files with high memory limit parameter (contributed by @bbuschkaemper).
• Fixed a potential issue in the linear stages of the clustering workflows when using a very high memory limit parameter (contributed by @bbuschkaemper).
• Fixed a bug that could cause an invalid error message or truncation when reading compressed files (present since v2.2.2).
• Added the option --symmetrize-evalue to maximize the evalue over both combinations of query and target.
• Fixed asymmetry of evalues in clustering workflows.
• Fixed a potential issue that could cause hanging instead of correct termination in case of an error.
• Fixed a dash character as file name not being interpreted as stdin or stdout.
• Fixed a wrong Illegal seek error message when using a stream as input.
• Added the --max-oid option to the greedy-vertex-cover workflow to interpret input edges as ordinal ids and set their maximum.
• Added the option --fpu-compat for exact inter-architecture equality of floating point computations when using --comp-based-stats 6.
• Fixed a bug that could cause a crash when using a custom seed shape of weight less than 7.