MeShClust
Release version

Requirements: g++ 4.9.1 or later, requires Homebrew on Mac OS X

Compilation using g++ (homebrew) and GNU Make on Mac OS X
CXX=g++-7 make

see: https://stackoverflow.com/questions/29057437/compile-openmp-programs-with-gcc-compiler-on-os-x-yosemite


Linux/Unix compilation:
make

If you find this tool helpful, please cite:

James, Benjamin T. et al. (2018), MeShClust: an intelligent tool for clustering DNA sequences. Nucleic Acids Research, gky315.

Usage: bin/meshclust *.fasta [--id 0.90] [--kmer 3] [--delta 5] [--output output.clstr] [--iterations 20] [--align] [--sample 3000] [--pivot 40] [--threads TMAX]

The most important parameter, --id, controls the identity of the sequences.
    If the identity is below 60%, alignment is automatically used instead of k-mer measures.
    However, alignment can be forced with the --align parameter.

--kmer decides the size of the kmers. It is by default automatically decided by average sequence length,
       but if provided, MeShClust can speed up a little by not having to find the largest sequence length.
       Increasing kmer size can increase accuracy, but increases memory consumption fourfold.

--delta decides how many clusters are looked around in the final clustering stage.
	Increasing it creates more accuracy, but takes more time.

--output specifies the output file, in CD-HIT's CLSTR format

--iterations specifies how many iterations in the final stage of merging are done until convergence.

--align forces alignment to be used, which can be much slower than k-mer features, but is
	more accurate than using k-mer features to guess alignment.

--threads sets the number of threads to be used. By default OpenMP uses the number of available cores
	  on your machine, but this parameter overwrites that.

--sample selects the total number of sample pairs of sequences used for both training and testing.
	 1500 is the default value.

--pivot selects the maximum number of pairs selected from one pivot sequence. Increasing this means
	less pivots are available, but more pairs are selected for one sequence, which can lead to
	higher training accuracy. The default value is 40.

If the argument is not listed here, it is interpreted as an input file.


License

Academic use: The software is provided as-is under the GNU GPLv3.
Any restrictions to use for-profit or non-academics: License needed.