Development manual for vsearch 2.32.0, built from the dev branch: it describes changes that are not released yet. The manual for the current release is published separately.

NAME

vsearch --uchime_ref — detect chimeras using a reference database

SYNOPSIS

vsearch --uchime_ref fastafile (--borderline | --chimeras | --nonchimeras | --uchimealns | --uchimeout) filename --db dbfile [options]

DESCRIPTION

The vsearch command --uchime_ref detects chimeric sequences present in the fasta-formatted fastafile by comparing them against a reference database of chimera-free sequences (option --db). Sequences are compared on their plus strand only; --strand both is not supported by --uchime_ref and is rejected.

Chimera detection is based on a scoring function controlled by five options: --dn, --mindiffs, --mindiv, --minh, and --xn. The algorithm identifies candidate chimeras by finding three-way alignments where a query sequence can be modelled as a mosaic of two parent sequences from the reference database.

Candidate parents are not gathered by comparing the query to every reference sequence. The query is split into four parts, each part is searched separately, and the four most similar sequences found for each part are kept, so the best pair of parents is chosen among at most sixteen candidates. That inner search runs with fixed limits of its own, which no option exposes. A reference matching one part very well but the whole query poorly can therefore crowd out the best overall parent, and candidates that tie are kept in the order they occur in the reference database, so reordering that file can change which parents are reported.

Chimeras can only be detected if their parents, or sufficiently close relatives, are present in the reference database. Unlike the de novo methods, --uchime_ref does not require abundance annotations. Multithreading is supported: queries are distributed over the available threads, so the order of the entries written to --borderline, --chimeras, --nonchimeras, --uchimealns and --uchimeout may vary from run to run when more than one thread is used. The results themselves do not depend on the thread count.

Both --db and at least one output option must be specified.

See also --uchime_denovo, --uchime2_denovo, and --uchime3_denovo for de novo chimera detection without a reference database.

OPTIONS

mandatory options

--uchime_ref fastafile: Detect chimeras in the fasta-formatted fastafile using a reference database.
--db filename: Read chimera-free reference sequences from filename. Chimeras cannot be detected if their parents, or sufficiently close relatives, are not present in the database. filename must refer to a fasta file, a fastq file (quality values are ignored), or a UDB file. If a UDB file is used, it should be created using the --makeudb_usearch command with the --dbmask dust option. --db accepts - to read the database from standard input, as well as an explicit stream path such as /dev/stdin, a named pipe, or a process substitution. The query and the database cannot both be -, however, as they would compete for the same standard input; give at least one of them an explicit path.

At least one of the following output options must be specified:

--borderline filename

Write borderline chimeric sequences to filename, in fasta format. Borderline chimeric sequences are sequences with a chimera score high enough to be suspicious, but not sufficiently different from their closest parent to be classified as chimeric.

--chimeras filename

Write chimeric sequences to filename, in fasta format.

--nonchimeras filename

Write non-chimeric sequences to filename, in fasta format.

--uchimealns filename

Write three-way global alignments (parentA, parentB, chimera) to filename in a human-readable format. All sequences are converted to upper case before alignment. Lower case letters indicate disagreement in the alignment. Use --alignwidth to modify the alignment width. The alignment spans the parents over their full length, so a parent region reaching beyond the query is shown as a run of gaps on the query line rather than trimmed away.

--uchimeout filename

Write chimera detection results to filename using an 18-field, tab-separated uchime-like format. Use --uchimeout5 for a format compatible with usearch version 5 and earlier. The 18 fields are:

score: higher score means a more likely chimeric alignment.
Q: query sequence label.
A: parent A sequence label.
B: parent B sequence label.
T: top parent sequence label (parent most similar to the query).
idQM: percentage of similarity between query (Q) and the model 13) constructed as a part of parent A and a part of parent B.
idQA: percentage of similarity between query (Q) and parent A.
idQB: percentage of similarity between query (Q) and parent B.
idAB: percentage of similarity between parent A and parent B.
idQT: percentage of similarity between query (Q) and top parent (T).
LY: yes votes in the left part of the model.
LN: no votes in the left part of the model.
LA: abstain votes in the left part of the model.
RY: yes votes in the right part of the model.
RN: no votes in the right part of the model.
RA: abstain votes in the right part of the model.
div: divergence, defined as (idQM - idQT).
YN: query is chimeric (Y), not chimeric (N), or borderline (?).

core options

--dbmask none|dust|soft: Mask regions in the database sequences using the dust method or the soft method, or none to suppress masking. See vsearch-fastx_mask(1) for more details. Warning, when using soft masking, search commands become case sensitive: masking excludes masked regions from the k-mer index used to select candidate targets (the pairwise alignment itself always ignores case). A database sequence with no unmasked stretch of at least the word length contributes no k-mer to the index and is never selected as a candidate target; it can still be reported when the query itself samples no k-mer, as such a query is compared against every database sequence (see --qmask). vsearch warns when there are such sequences, giving how many. The default is to mask using dust.
--dn strictly positive real number: Set the pseudo-count prior on the number of ‘no’ votes, corresponding to the parameter n in the chimera scoring function. Default value is 1.4. Increasing --dn reduces the likelihood of tagging a sequence as a chimera (fewer false positives, but also more false negatives).
--mindiffs positive integer: Set the minimum number of differences per segment. Default value is 3.
--mindiv real: Set the minimum divergence from the closest parent. Default value is 0.8.
--minh real: Set the minimum chimera score (h). Increasing this value tends to reduce the number of false positives and to decrease sensitivity. Default value is 0.28. Accepted values are strictly greater than 0.0; values above 1.0 are accepted but uncommon.
--self: Ignore a reference sequence when its label matches the label of the query sequence. Useful to estimate the false-positive rate in reference sequences.
--selfid: Ignore a reference sequence when its nucleotide sequence is strictly identical to the nucleotide sequence of the query.
--strand plus: Check the plus strand only. This is the only accepted value: --uchime_ref rejects --strand both with a fatal error. The option is accepted for compatibility with usearch. Keywords are case-insensitive.
--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 value 0 is also accepted and, like the default, uses all available cores; decimal values are truncated to their integer part. On Linux, “available” accounts for the CPU affinity mask and the cgroup CPU quota of the running process, so a job confined by taskset(1), Slurm, Docker or Kubernetes launches one thread per core it was actually granted, rather than one per core the machine has.
--uchimeout5: When using --uchimeout, write chimera detection results using a 17-field, tab-separated uchime-like format. This drops the 5th field (top parent T) from the standard --uchimeout format, for compatibility with usearch version 5 and earlier versions.
--xn real number strictly greater than 1.0: Set the weight of ‘no’ votes, corresponding to the parameter beta in the chimera scoring function. Default value is 8.0. Increasing --xn reduces the likelihood of tagging a sequence as a chimera (less false positives, but also more false negatives). Decreasing --xn reduces false negative, but increases false positives.

secondary options

--alignwidth positive integer

Set the width of the three-way alignments written with --uchimealns. Default width is 80 nucleotides. Set to zero (0) to suppress folding.

--fasta_score

Add the chimera score to the sequence headers in the fasta output files for chimeras, non-chimeras, and borderline sequences, using the format ;uchime_ref=float (appended at the end of the header).

--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.

--hardmask

Replace masked nucleotides with Ns, rather than lowercasing them.

This is also the only masking option that reaches the alignment. Soft and dust masking (see --qmask and --dbmask) only keep masked words out of the k-mer pre-filter that selects candidates; the masked region is still aligned and scored like any other. An N, on the other hand, scores zero and counts as a matching column, so hard masking makes a masked region match whatever it is aligned against, and the reported identity can only go up. Add --n_mismatch to count those columns as mismatches instead.

--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’.

--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. Messages are written in addition to the standard error, not instead of it; use --quiet to silence the standard error copy (see the --quiet entry for what it spares).

--maxseqlength positive integer

Discard sequences longer than positive integer (50,000 nucleotides by default). The value must not exceed 2,147,481,646 (INT_MAX minus 2,001). When that value is smaller than the effective --minseqlength value (whose default is command-specific), no sequence can pass the length filter: vsearch issues a warning.

--minseqlength non-negative integer

Discard sequences shorter than non-negative integer (1 nucleotide by default). A value of 0 retains empty sequences. When that value is larger than the --maxseqlength value, no sequence can pass the length filter: vsearch issues a warning.

--no_progress

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

--notrunclabels

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

--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: masking excludes masked regions from the k-mer pre-filter that selects candidate targets (the pairwise alignment itself always ignores case). A query with no unmasked stretch of at least the word length samples no k-mers and is therefore compared against every database sequence. The default is to mask using dust.

--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.

--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. When nothing is left after truncation — an empty string, or one starting with ‘;’ or a blank character — vsearch issues a warning and writes a bare ;sample= annotation.

--sizeout

Add abundance annotations to sequence headers when writing fasta or fastq files. Add the pattern ;size=integer. Existing ;size= annotations are reported unchanged; entries without one receive ;size=1. For this command --sizein is not needed: abundance annotations are always parsed from the input headers.

--xee

Strip expected error (ee) annotations from sequence headers when writing fasta or fastq files. Search for the pattern [>@;]ee=float[;]. Expected error annotations are added by the synonymous options --fastq_eeout and --eeout described in vsearch-fastx_filter(1).

--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.

pairwise alignment options

These options modify the parameters of the pairwise alignment model. Modify with caution.

--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.
--match integer: Set the score assigned to a match (i.e. equivalent nucleotides) in pairwise alignments. The default value is 2. Accepted values range from -32767 to +32767.
--mismatch integer: Set the score assigned to a mismatch (i.e. different nucleotides) in pairwise alignments. The default value is -4. Accepted values range from -32767 to +32767.

ignored options

--abskew real: Set the minimum abundance skew ratio between a chimera and its potential parent. The assumption is that chimeras appear later in the PCR amplification process and are therefore less abundant than their parents. Any positive value equal or greater than 1.0 can be used. Default is 2.0, which means that the parents should be at least 2 times more abundant than the chimera.; Ignored by --uchime_ref: the abundance-skew rule applies to the de novo commands only.
--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.; Has no effect with --uchime_ref: abundance annotations are always parsed (entries without annotations count as 1) and feed the summary counts and --sizeout.

EXAMPLES

Detect chimeras using a reference database and write non-chimeric sequences to a file:

vsearch \
    --uchime_ref amplicons.fasta \
    --db silva_138_db.fasta \
    --nonchimeras clean.fasta

Use multiple threads and write a detailed tab-separated report:

vsearch \
    --uchime_ref amplicons.fasta \
    --db silva_138_db.fasta \
    --threads 4 \
    --nonchimeras clean.fasta \
    --uchimeout chimera_report.tsv

Write all three output categories (chimeras, non-chimeras and borderline entries):

vsearch \
    --uchime_ref amplicons.fasta \
    --db silva_138_db.fasta \
    --nonchimeras clean.fasta \
    --chimeras chimeras.fasta \
    --borderline borderline.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.

These manual pages are also published online at https://torognes.github.io/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.