NAME

vsearch \-\-cluster_smallmem — clusterize pre-sorted sequences

SYNOPSIS

DESCRIPTION

The vsearch command --cluster_smallmem groups the fasta sequences in fastafile into clusters using a greedy, heuristic, centroid-based algorithm. Unlike --cluster_fast and --cluster_size, this command does not sort the input sequences before clustering; sequences are expected to be sorted by decreasing length. Use --usersort to override this requirement and allow sequences in any order. At least one output option must be specified.

For each query sequence (in input order), vsearch compares it to all existing cluster centroids. If the query is similar enough to a centroid (as determined by --id), it is assigned to that cluster; otherwise, a new cluster is seeded with the query as its centroid.

Sequences are compared using global pairwise alignment (Needleman-Wunsch). The number of comparisons is limited by --maxaccepts and --maxrejects.

--cluster_fast (see vsearch-cluster_fast(1)) and --cluster_size (see vsearch-cluster_size(1)) perform the same clustering but automatically sort the input sequences beforehand.

See vsearch-fasta(5) for a description of the input format.

OPTIONS

mandatory options

--cluster_smallmem fastafile
Read and clusterize fasta sequences from fastafile. Sequences must be sorted by decreasing length, unless --usersort is used. This option is mandatory.

--id real
Reject a sequence if its pairwise identity with the cluster centroid is lower than real (value ranging from 0.0 to 1.0 included). The pairwise identity is defined by default as (matching columns) / (alignment length - terminal gaps). That definition can be modified with --iddef.

core options

--centroid_sizeout
Add centroid abundance annotations to sequence headers when writing fasta files. Add the pattern ;centroid_size=integer.

--centroids filename
Write cluster centroid sequences to filename, in fasta format. The centroid is the sequence that seeded the cluster (i.e. the first sequence of the cluster).

--clusterout_id
Add cluster identifier information to the output files when using --centroids, --consout, and --profile.

--clusterout_sort
Sort some output files by decreasing abundance instead of input order. Applies to --consout, --msaout, --profile, --centroids, and --uc. For --uc, the sorting applies only to the centroid information part (the C lines).

--clusters string
Write each cluster to a separate fasta file using string as a prefix. A ticker (0, 1, 2, etc.) is appended to construct the path and filename for each cluster.

--consout filename
Write cluster consensus sequences to filename. For each cluster, a center-star multiple sequence alignment is computed with the centroid as the center, using a fast algorithm. A consensus sequence is constructed by taking the majority symbol (nucleotide or gap) from each column of the alignment. Columns containing a majority of gaps are skipped, except for terminal gaps. If --sizein is specified, sequence abundances are taken into account.

--iddef 0|1|2|3|4
Change the pairwise identity definition used with --id. Accepted values are:

CD-HIT definition: (matching columns) / (shortest sequence length).
edit distance: (matching columns) / (alignment length).
edit distance excluding terminal gaps (default definition for --id).
Marine Biological Lab definition, counting each gap opening (internal or terminal) as a single mismatch, whether or not the gap was extended: 1.0 - [(mismatches + gap openings)/(longest sequence length)].
BLAST definition, equivalent to --iddef 1 for global pairwise alignments.

--maxaccepts positive integer
Set the maximum number of matching target sequences to accept before stopping the search for a given query. The default value is 1. Use together with --maxrejects. If both --maxaccepts and --maxrejects are set to 0, the complete database is searched.

--maxrejects positive integer
Set the maximum number of non-matching target sequences to consider before stopping the search for a given query. The default value is 32. Use together with --maxaccepts. If both --maxaccepts and --maxrejects are set to 0, the complete database is searched.

--msaout filename
Write a multiple sequence alignment and a consensus sequence for each cluster to filename, in fasta format. vsearch computes center-star multiple sequence alignments using a fast method whose accuracy can decrease at low pairwise identity thresholds. The consensus sequence is constructed by taking the majority symbol (nucleotide or gap) from each column of the alignment. Columns containing a majority of gaps are skipped, except for terminal gaps. If --sizein is specified, sequence abundances are taken into account when computing the consensus.

--profile filename
Write a sequence profile to filename, reporting the frequency of each nucleotide at each position in the multiple alignment for each cluster. A FASTA-like header line precedes the profile information for each cluster. The data is tab-separated with eight columns: position (0-based), consensus nucleotide, number of As, Cs, Gs, Ts or Us, gap symbols, and total number of ambiguous nucleotide symbols (B, D, H, K, M, N, R, S, Y, V or W). If --sizein is specified, sequence abundances are taken into account.

--qmask none|dust|soft
Mask regions in query sequences using the dust method or the soft method, or none to suppress masking. Values are case-insensitive, so DUST, Dust, and dust are all accepted. See vsearch-fastx_mask(1) for more details. Warning, when using soft masking, search commands become case sensitive. The default is to mask using dust.

--sizein
Use the abundance annotations present in sequence headers when reading fasta or fastq file. Search for the pattern [>@;]size=integer[;]. Entries without abundance annotations are silently assumed to be of size=1.

--sizeorder
When a sequence is close to two or more centroids within the distance specified by --id, resolve the ambiguity by assigning it to the centroid with the highest abundance, rather than the closest one. Only takes effect when --maxaccepts is greater than one. This option enables abundance-based greedy clustering (AGC), as opposed to the default distance-based greedy clustering (DGC).

--sizeout
Add abundance annotations to sequence headers when writing fasta or fastq files. Add the pattern ;size=integer. If option --sizein is not used, abundance values are set to 1 for all entries. If --sizein is used, existing abundance annotations are simply reported to output files.

--strand plus|both
Check the plus strand only (default), or check both strands when comparing sequences.

--uc filename
Write results to filename in a tab-separated uclust-like format with 10 columns. Three record types are used per row: cluster seeds (S), hits (H), and cluster summaries (C). Columns are:

record type (S, H, or C);
cluster number (zero-based);
centroid length (S), query length (H), or cluster size (C);
percent identity with centroid (H), or * (S, C);
match orientation + or - (H), or * (S, C);
not used; always 0 (H) or * (S, C);
not used; always 0 (H) or * (S, C);
CIGAR alignment string (H), or * (S, C); see vsearch-cigar(5);
query label (H), or centroid label (S, C);
centroid label (H), or * (S, C).

--usersort
When using --cluster_smallmem, allow input sequences in any order, rather than requiring them to be sorted by decreasing length.

secondary options

--alnout filename
Write pairwise global alignments to filename in a human-readable format. Use --rowlen to set the alignment line width.

--biomout filename
Write an OTU table to filename in the biom version 1.0 JSON file format. The OTUs are represented by the cluster centroids. Sample identifiers are extracted from sequence headers (;sample=abc123; or ;barcodelabel=abc123; patterns, or the initial part of the header). OTU identifiers are extracted from centroid headers (;otu=def789; pattern, or the initial part of the header, or via relabelling options). Taxonomy information is extracted from centroid headers (;tax=...; pattern) if available.

--blast6out filename
Write results to filename using a BLAST-like tab-separated format with twelve fields per query-target match: query label, target label, percentage identity, alignment length, mismatches, gap openings, query start, query end, target start, target end, expectation value (always -1), and bit score (always 0). If --output_no_hits is used, non-matching queries are also written. Note that vsearch uses global pairwise alignments, not BLAST’s seed-and-extend algorithm.

--bzip2_decompress
Specify that the input pipe is streaming data compressed using Huffman coding. See bzip2(1) for more details. This option is not needed when reading from a regular file compressed with bzip2.

--fastapairs filename
Write pairwise alignments of query and target sequences to filename, in fasta format.

--fasta_width positive integer
Set the maximal width of sequences when writing fasta files. Longer sequences are folded and written on several lines. Default width is 80 nucleotides. Set to zero (0) to suppress folding.

--gapext string
Set penalties for a gap extension. See vsearch-pairwise_alignment_parameters(7) for a complete description of the gap penalty declaration system. By default, the penalty is set to 2 for extending internal gaps and to 1 for extending terminal gaps, in both query and target sequences.

--gapopen string
Set penalties for a gap opening. See vsearch-pairwise_alignment_parameters(7) for a complete description of the gap penalty declaration system. By default, the penalty is set to 20 for opening internal gaps and to 2 for opening terminal gaps, in both query and target sequences.

--gzip_decompress
Specify that the input pipe is streaming data compressed using Lempel-Ziv coding. See gzip(1) for more details. This option is not needed when reading from a regular file compressed with gzip.

--hardmask
Replace masked nucleotides with Ns.

--idprefix positive integer
Reject the sequence match if the first integer nucleotides of the target do not match the query.

--idsuffix positive integer
Reject the sequence match if the last integer nucleotides of the target do not match the query.

--label_suffix string
Add the suffix string to sequence headers when writing fasta or fastq files. For example, with --label_suffix ";status=healthy", sequence header ‘>seq1’ becomes ‘>seq1;status=healthy’.

--leftjust
Reject the sequence match if the pairwise alignment begins with gaps.

--lengthout
Add a sequence length annotation (;length=integer) to each sequence header when writing fasta or fastq files.

--log filename
Write messages to filename. Messages include program version, start and finish times, elapsed time, amount of memory available, maximum amount of memory consumed, number of cores and command line options, and if need be, command-specific informational messages, warnings, and errors.

--match integer
Set the score assigned to a match (i.e. equivalent nucleotides) in pairwise alignments. The default value is 2.

--matched filename
Write query sequences matching a target sequence to filename, in fasta format.

--maxdiffs positive integer
Reject the sequence match if the alignment contains more than integer substitutions, insertions, or deletions.

--maxgaps positive integer
Reject the sequence match if the alignment contains more than integer insertions or deletions.

--maxhits non-negative integer
Set the maximum number of hits to report once the search is terminated for a given query; hits are sorted by decreasing identity. Unlimited by default, or when the argument is zero. When searching both strands, --maxhits controls the total number of hits reported per query across both strands.

--maxid real
Reject the sequence match if the pairwise identity between the two sequences is greater than real.

--maxqsize positive integer
Reject query sequences with an abundance greater than integer.

--maxqt real
Reject the sequence match if the query/target sequence length ratio is greater than real.

--maxseqlength positive integer
Discard sequences longer than positive integer (50,000 nucleotides by default).

--maxsizeratio real
Reject the sequence match if the query/target abundance ratio is greater than real.

--maxsl real
Reject the sequence match if the shorter/longer sequence length ratio is greater than real.

--maxsubs positive integer
Reject the sequence match if the pairwise alignment contains more than integer substitutions.

--mid real
Reject the sequence match if the pairwise identity, computed ignoring all gaps (internal and terminal), is lower than real.

--mincols positive integer
Reject the sequence match if the alignment length is shorter than integer columns.

--minseqlength positive integer
Discard sequences shorter than positive integer (32 nucleotides by default).

--minqt real
Reject the sequence match if the query/target sequence length ratio is lower than real.

--minsizeratio real
Reject the sequence match if the query/target abundance ratio is lower than real.

--minsl real
Reject the sequence match if the shorter/longer sequence length ratio is lower than real.

--mintsize positive integer
Reject target sequences with an abundance lower than integer.

--minwordmatches non-negative integer
Set the minimum number of k-mer matches required for a sequence to be considered further. The default value is 12 for the default word length of 8. If the argument is 0, no word matches are required.

--mismatch integer
Set the score assigned to a mismatch (i.e. different nucleotides) in pairwise alignments. The default value is -4.

--mothur_shared_out filename
Write an OTU table to filename in the mothur ‘shared’ tab-separated plain text format. The first line starts with label, group and numOtus, followed by all OTU identifiers. Each subsequent line starts with vsearch, the sample identifier, the total number of OTUs, and the abundance of each OTU in that sample. Sample and OTU identifiers are extracted from FASTA headers. OTUs are represented by the cluster centroids.

--n_mismatch
Count alignments of nucleotides against Ns as mismatches.

--no_progress
Suppress the gradually increasing progress indicator normally written to the standard error stderr(3).

--notmatched filename
Write the sequences that were not extracted to filename, in fasta format.

--notrunclabels
Retain whole sequence headers in output files. By default, vsearch truncates sequence headers at first space or tabulation. This option suppresses truncation.

--otutabout filename
Write an OTU table to filename in a classic tab-separated plain text format. The first line starts with #OTU ID followed by sample identifiers. Each subsequent line starts with the OTU identifier followed by the abundances in each sample. Sample and OTU identifiers are extracted from FASTA headers (see --sample). OTUs are represented by the cluster centroids. A taxonomy column is appended if taxonomy information is available for any OTU.

--output_no_hits
Write both matching and non-matching queries to --alnout, --blast6out, --samout, or --userout output files. Non-matching queries are labelled ‘No hits’ in --alnout files.

--query_cov real
Reject the sequence match if the fraction of the query aligned to the target is lower than real (value ranging from 0.0 to 1.0 included). Query coverage is computed as (matches + mismatches) / query sequence length, not counting internal or terminal gaps.

--quiet
Suppress messages to the standard output stdout(3) and standard error stderr(3), except for warnings and error messages.

--relabel string
Replace sequence headers with the prefix string and a ticker (1, 2, 3, etc.). For example, with --relabel "cluster:", the first sequence header becomes ‘>cluster:1’, the second sequence header becomes ‘>cluster:2’, and so on. To retain annotations, use their corresponding options (--lengthout, --eeout, and --sizeout). Use --relabel_keep to also retain old sequence identifiers.

--relabel_keep
Retain old sequence identifiers by including them at the end of the new headers, after a space.

--relabel_md5
Replace each sequence header with the MD5 digest derived from the sequence itself. The sequence is converted to upper case, and each ‘U’ is replaced with a ‘T’ before computation of the digest. The MD5 digest is a 128-bit value (16 bytes), represented using a string of 32 ASCII characters. Each pair of characters encodes an hexadecimal value, ranging from x00 to xff. See md5(3) for more details, and --relabel_sha1 for an alternative hashing algorithm. To retain annotations, use their corresponding options (--lengthout, --eeout, and --sizeout). Use --relabel_keep to also retain old sequence identifiers.

--relabel_self
Replace each sequence header with the sequence itself. To retain annotations, use their corresponding options (--lengthout, --eeout, and --sizeout). Use --relabel_keep to also retain old sequence identifiers.

--relabel_sha1
Replace each sequence header with the SHA1 digest derived from the sequence itself. The sequence is converted to upper case, and each ‘U’ is replaced with a ‘T’ before computation of the digest. The SHA1 digest is a 160-bit value (20 bytes), represented using a string of 40 ASCII characters. Each pair of characters encodes an hexadecimal value, ranging from x00 to xff. See sha1(3) for more details, and --relabel_md5 for an alternative hashing algorithm. To retain annotations, use their corresponding options (--lengthout, --eeout, and --sizeout). Use --relabel_keep to also retain old sequence identifiers.

--rightjust
Reject the sequence match if the pairwise alignment ends with gaps.

--rowlen positive integer
Set the width of alignment lines in --alnout output. The default value is 64. Set to 0 to disable line wrapping.

--samheader
Include header lines (@HD, @SQ, @PG) in the SAM file produced by --samout. By default, no header lines are written.

--samout filename
Write alignment results to filename in the SAM format, see vsearch-sam(5). Use --samheader to include header lines. Each non-header line is a SAM record representing either a query-target alignment or the absence of a match. The alignment column of each record uses the CIGAR format, see vsearch-cigar(5).

--sample string
Add the given sample identifier string to sequence headers when writing fasta or fastq files. For instance, if string is ‘ABC’, the text ;sample=ABC will be added to the headers. string is silently truncated at the first ‘;’ or whitespace character (space, tab, newline, carriage return, vertical tab or form feed), so such characters should not be used in string. Other characters (alphabetical, numerical and punctuations) are accepted.

--self
Reject the sequence match if the query and target sequence labels are identical.

--selfid
Reject the sequence match if the query and target sequences are strictly identical.

--qsegout filename
Write the aligned part of each query sequence to filename, in fasta format.

--target_cov real
Reject the sequence match if the fraction of the target sequence aligned to the query is lower than real. Target coverage is computed as (matches + mismatches) / target sequence length, not counting internal or terminal gaps.

--threads positive integer
Set the number of computation threads to use, from 1 to 1024. The number of threads should not exceed the number of available CPU cores. The default is to use all available cores.

--top_hits_only
Report only the hits with the highest pairwise identity for each query.

--tsegout filename
Write the aligned part of each target sequence to filename, in fasta format.

--userfields string
Select and order the fields written to --userout output. Fields are separated by + (e.g. query+target+id). See vsearch-userfields(7) for a complete description of all available fields.

--userout filename
Write user-defined tab-separated output to filename. Select and order the fields with --userfields.

--weak_id real
Report hits with a pairwise identity of at least real, without stopping the search. Unlike --id, weak hits do not count toward --maxaccepts but do count toward --maxrejects. The value of real must be smaller than the value specified with --id.

--wordlength positive integer
Set the length of words (i.e. k-mers) used for sequence comparisons. The range of possible values goes from 3 to 15 (default value is 12). Longer words may reduce the sensitivity/recall for weak similarities, but can increase precision. On the other hand, shorter words may increase sensitivity or recall, but may reduce precision.

Computation time generally increases with shorter words and decreases with longer words, but it increases again for very long words. Memory requirements for a part of the index increase by a factor of 4 each time word length increases by one nucleotide, and this generally becomes significant for words longer than 12.

--xlength
Strip sequence length annotations from sequence headers when writing fasta or fastq files. Search for the pattern [>@;]length=integer[;]. Sequence length annotations are added by the --lengthout option.

--xsize
Strip abundance annotations from sequence headers when writing fasta or fastq files. Search for the pattern [>@;]size=integer[;]. Abundance annotations are added by the --sizeout option.

ignored options

--band positive integer
This option is ignored. It is provided for compatibility with usearch.

--cons_truncate
This option is ignored. A warning is issued.

--fulldp
This option is ignored. It is provided for compatibility with usearch. vsearch always uses a full dynamic programming algorithm (Needleman-Wunsch).

--hspw positive integer
This option is ignored. It is provided for compatibility with usearch.

--minhsp positive integer
This option is ignored. It is provided for compatibility with usearch.

--pattern string
This option is ignored. It is provided for compatibility with usearch.

--slots positive integer
This option is ignored. It is provided for compatibility with usearch.

--xdrop_nw positive integer
This option is ignored. It is provided for compatibility with usearch.

EXAMPLES

Cluster sequences at 97% identity, assuming the input is already sorted by decreasing length:

vsearch \
    --cluster_smallmem sequences.fasta \
    --id 0.97 \
    --sizein \
    --sizeout \
    --centroids centroids.fasta

Cluster sequences in any order by using --usersort:

vsearch \
    --cluster_smallmem sequences.fasta \
    --id 0.97 \
    --usersort \
    --sizein \
    --sizeout \
    --centroids centroids.fasta

CITATION

Rognes T, Flouri T, Nichols B, Quince C, Mahé F. (2016) VSEARCH: a versatile open source tool for metagenomics. PeerJ 4:e2584 doi: 10.7717/peerj.2584

REPORTING BUGS

Submit suggestions and bug-reports at https://github.com/torognes/vsearch/issues, send a pull request on https://github.com/torognes/vsearch, or compose a friendly or curmudgeont e-mail to Torbjørn Rognes (torognes@ifi.uio.no).

AVAILABILITY

Source code and binaries are available at https://github.com/torognes/vsearch.

COPYRIGHT

Contact: Torbjørn Rognes torognes@ifi.uio.no, Department of Informatics, University of Oslo, PO Box 1080 Blindern, NO-0316 Oslo, Norway

This software is dual-licensed and available under a choice of one of two licenses, either under the terms of the GNU General Public License version 3 or the BSD 2-Clause License.

GNU General Public License version 3

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.

The BSD 2-Clause License

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

ACKNOWLEDGMENTS

We would like to thank the authors of the following projects for making their source code available:

vsearch includes code from Google’s CityHash project by Geoff Pike and Jyrki Alakuijala, providing some excellent hash functions available under a MIT license.
vsearch includes code derived from Tatusov and Lipman’s DUST program that is in the public domain.
vsearch includes public domain code written by Alexander Peslyak for the MD5 message digest algorithm.
vsearch includes public domain code written by Steve Reid and others for the SHA1 message digest algorithm.
vsearch binaries may include code from the zlib library, copyright Jean-Loup Gailly and Mark Adler.
vsearch binaries may include code from the bzip2 library, copyright Julian R. Seward.

vsearch

Versatile open-source tool for microbiome analysis