mode_query.txt

SYNOPSIS

    metacache query <database>

    metacache query <database> <sequence file/directory>... [OPTION]...

    metacache query <database> [OPTION]... <sequence file/directory>...


DESCRIPTION

    Map sequences (short reads, long reads, genome fragments, ...)
    to their most likely taxon of origin.


BASIC PARAMETERS

    <database>        database file name;
                      A MetaCache database contains taxonomic information and
                      min-hash signatures of reference sequences (complete
                      genomes, scaffolds, contigs, ...).

    <sequence file/directory>...
                      FASTA or FASTQ files containing genomic sequences (short
                      reads, long reads, contigs, complete genomes, ...) that
                      shall be classified.
                      * If directory names are given, they will be searched for
                      sequence files (at most 10 levels deep).
                      * If no input filenames or directories are given,
                      MetaCache will run in interactive query mode. This can be
                      used to load the database into memory only once and then
                      query it multiple times with different query options.


MAPPING RESULTS OUTPUT

    -out <file>       Redirect output to file <file>.
                      If not specified, output will be written to stdout. If
                      more than one input file was given all output will be
                      concatenated into one file.


    -split-out <file> Generate output and statistics for each input file
                      separately. For each input file <in> an output file with
                      name <file>_<in> will be written.


PAIRED-END READ HANDLING

    -pairfiles        Interleave paired-end reads from two consecutive files, so
                      that the nth read from file m and the nth read from file
                      m+1 will be treated as a pair. If more than two files are
                      provided, their names will be sorted before processing.
                      Thus, the order defined by the filenames determines the
                      pairing not the order in which they were given in the
                      command line.


    -pairseq          Two consecutive sequences (1+2, 3+4, ...) from each file
                      will be treated as paired-end reads.


    -insertsize <#>   Maximum insert size to consider.
                      default: sum of lengths of the individual reads


READ FILTERS

    -min-readlen <#>  Minimum length of reads (in bp). Reads shorter than the
                      given number will not be classified.
                      default: no limit

    -max-readlen <#>  Maximum length of reads (in bp). Reads longer than the
                      given number will not be classified.
                      default: no limit


CLASSIFICATION

    -lowest <rank>    Do not classify on ranks below <rank>
                      (Valid values: sequence, form, variety, subspecies,
                      species, subgenus, genus, subtribe, tribe, subfamily,
                      family, suborder, order, subclass, class, subphylum,
                      phylum, subkingdom, kingdom, domain)
                      default: sequence

    -highest <rank>   Do not classify on ranks above <rank>
                      (Valid values: sequence, form, variety, subspecies,
                      species, subgenus, genus, subtribe, tribe, subfamily,
                      family, suborder, order, subclass, class, subphylum,
                      phylum, subkingdom, kingdom, domain)
                      default: domain

    -hitmin <t>       Sets classification threshhold to <t>.
                      A read will not be classified if less than t features from
                      the database match. Higher values will increase precision
                      at the expense of sensitivity.
                      default: 0

    -hitdiff <d>      Sets candidate LCA threshhold to <d> percent.
                      Influences if only candidate with the most hits will be
                      used as classification result or if taxa of other
                      candidates will be considered.
                      All candidate (taxa) will be included that have at least
                      d% as many hits above the hit-min threshold as the
                      candidate with the most hits.
                      default: 100

    -maxcand <#>      maximum number of reference taxon candidates to consider
                      for each query;
                      A large value can significantly decrease the querying
                      speed!.
                      default: 2

    -cov-percentile <p>
                      Remove the p-th percentile of hit reference sequences with
                      the lowest coverage. Classification is done using only the
                      remaining reference sequences. This can help to reduce
                      false positives, especially whenyour input data has a high
                      sequencing coverage.
                      This feature decreases the querying speed!
                      default: off


GENERAL OUTPUT FORMATTING

    -no-summary       Dont't show result summary & mapping statistics at the end
                      of the mapping output
                      default: off

    -no-query-params  Don't show query settings at the beginning of the mapping
                      output
                      default: off

    -no-err           Suppress all error messages.
                      default: off


CLASSIFICATION RESULT FORMATTING

    -no-map           Don't report classification for each individual query
                      sequence; show summaries only (useful for quick tests).
                      default: off

    -mapped-only      Don't list unclassified reads/read pairs.
                      default: off

    -taxids           Print taxon ids in addition to taxon names.
                      default: off

    -taxids-only      Print taxon ids instead of taxon names.
                      default: off

    -omit-ranks       Do not print taxon rank names.
                      default: off

    -separate-cols    Prints *all* mapping information (rank, taxon name, taxon
                      ids) in separate columns (see option '-separator').
                      default: off

    -separator <text> Sets string that separates output columns.
                      default: '\t|\t'

    -comment <text>   Sets string that precedes comment (non-mapping) lines.
                      default: '# '

    -queryids         Show a unique id for each query.
                      Note that in paired-end mode a query is a pair of two read
                      sequences. This option will always be activated if option
                      '-hits-per-ref' is given.
                      default: off

    -lineage          Report complete lineage for per-read classification
                      starting with the lowest rank found/allowed and ending
                      with the highest rank allowed. See also options '-lowest'
                      and '-highest'.
                      default: off


ANALYSIS: ABUNDANCES

    -abundances <file>
                      Show absolute and relative abundance of each taxon.
                      If a valid filename is given, the list will be written to
                      this file.
                      default: off

    -abundance-per <rank>
                      Show absolute and relative abundances for each taxon on
                      one specific rank.
                      Classifications on higher ranks will be estimated by
                      distributing them down according to the relative
                      abundances of classifications on or below the given rank.
                      (Valid values: sequence, form, variety, subspecies,
                      species, subgenus, genus, subtribe, tribe, subfamily,
                      family, suborder, order, subclass, class, subphylum,
                      phylum, subkingdom, kingdom, domain)
                      If '-abundances <file>' was given, this list will be
                      printed to the same file.
                      default: off


ANALYSIS: RAW DATABASE HITS

    -tophits          For each query, print top feature hits in database.
                      default: off

    -allhits          For each query, print all feature hits in database.
                      default: off

    -locations        Show locations in candidate reference sequences.
                      Activates option '-tophits'.
                      default: off

    -hits-per-ref <file>
                      Shows a list of all hits for each reference sequence.
                      If this condensed list is all you need, you should
                      deactive the per-read mapping output with '-no-map'.
                      If a valid filename is given after '-hits-per-ref', the
                      list will be written to a separate file.
                      Option '-queryids' will be activated and the lowest
                      classification rank will be set to 'sequence'.
                      default: off


ANALYSIS: ALIGNMENTS

    -align            Show semi-global alignment to best candidate reference
                      sequence.
                      Original files of reference sequences must be available.
                      This feature decreases the querying speed!
                      default: off


ADVANCED: GROUND TRUTH BASED EVALUATION

    -ground-truth     Report correct query taxa if known.
                      Queries need to have either a 'taxid|<number>' entry in
                      their header or a sequence id that is also present in the
                      database.
                      This feature decreases the querying speed!
                      default: off

    -precision        Report precision & sensitivity by comparing query taxa
                      (ground truth) and mapped taxa.
                      Queries need to have either a 'taxid|<number>' entry in
                      their header or a sequence id that is also found in the
                      database.
                      This feature decreases the querying speed!
                      default: off

    -taxon-coverage   Report true/false positives and true/false negatives.This
                      option turns on '-precision', so ground truth data needs
                      to be available.
                      This feature decreases the querying speed!
                      default: off


ADVANCED: CUSTOM QUERY SKETCHING (SUBSAMPLING)

    -kmerlen <k>      number of nucleotides/characters in a k-mer
                      default: determined by database

    -sketchlen <s>    number of features (k-mer hashes) per sampling window
                      default: determined by database

    -winlen <w>       number of letters in each sampling window
                      default: determined by database

    -winstride <l>    distance between window starting positions
                      default: determined by database


ADVANCED: DATABASE MODIFICATION

    -max-locations-per-feature <#>
                      maximum number of reference sequence locations to be
                      stored per feature;
                      If the value is too high it will significantly impact
                      querying speed. Note that an upper hard limit is always
                      imposed by the data type used for the hash table bucket
                      size (set with compilation macro
                      '-DMC_LOCATION_LIST_SIZE_TYPE').
                      default: 254

    -remove-overpopulated-features
                      Removes all features that have reached the maximum allowed
                      amount of locations per feature. This can improve querying
                      speed and can be used to remove non-discriminative
                      features.
                      default: off
                      Not available in the GPU version.

    -remove-ambig-features <rank>
                      Removes all features that have more distinct reference
                      sequence on the given taxonomic rank than set by
                      '-max-ambig-per-feature'. This can decrease the database
                      size significantly at the expense of sensitivity. Note
                      that the lower the given taxonomic rank is, the more
                      pronounced the effect will be.
                      Valid values: sequence, form, variety, subspecies,
                      species, subgenus, genus, subtribe, tribe, subfamily,
                      family, suborder, order, subclass, class, subphylum,
                      phylum, subkingdom, kingdom, domain
                      default: off
                      Not available in the GPU version.

    -max-ambig-per-feature <#>
                      Maximum number of allowed different reference sequence
                      taxa per feature if option '-remove-ambig-features' is
                      used.
                      Not available in the GPU version.

    -max-load-fac <factor>
                      maximum hash table load factor;
                      This can be used to trade off larger memory consumption
                      for speed and vice versa. A lower load factor will improve
                      speed, a larger one will improve memory efficiency.
                      default: 0.800000
                      Not available in the GPU version.


ADVANCED: PERFORMANCE TUNING / TESTING

    -threads <#>      Sets the maximum number of parallel threads to use.default
                      (on this machine): 32


    -batch-size <#>   Process <#> many queries (reads or read pairs) per thread
                      at once.
                      default (on this machine): 4096

    -query-limit <#>  Classify at max. <#> queries (reads or read pairs) per
                      input file.
                      default: no limit


EXAMPLES

    Query all sequences in 'myreads.fna' against pre-built database 'refseq':
        metacache query refseq myreads.fna -out results.txt

    Query all sequences in multiple files against database 'refseq':
        metacache query refseq reads1.fna reads2.fna reads3.fna

    Query all sequence files in folder 'test' againgst database 'refseq':
        metacache query refseq test

    Query multiple files and folder contents against database 'refseq':
        metacache query refseq file1.fna folder1 file2.fna file3.fna folder2

    Perform a precision test and show all ranks for each classification result:
        metacache query refseq reads.fna -precision -allranks -out results.txt

    Load database in interactive query mode, then query multiple read batches
        metacache query refseq
        reads1.fa reads2.fa -pairfiles -insertsize 400
        reads3.fa -pairseq -insertsize 300
        reads4.fa -lineage


OUTPUT FORMAT

    MetaCache's default read mapping output format is:
    read_header | rank:taxon_name

    This will not be changed in the future to avoid breaking anyone's
    pipelines. Command line options won't change in the near future for the
    same reason. The following table shows some of the possible mapping
    layouts with their associated command line arguments:

    read mapping layout                          command line arguments
    ---------------------------------------      ---------------------------------
    read_header | taxon_id                       -taxids-only -omit-ranks
    read_header | taxon_name                     -omit-ranks
    read_header | taxon_name(taxon_id)           -taxids -omit-ranks
    read_header | taxon_name | taxon_id          -taxids -omit-ranks -separate-cols
    read_header | rank:taxon_id                  -taxids-only
    read_header | rank:taxon_name
    read_header | rank:taxon_name(taxon_id)      -taxids
    read_header | rank | taxon_id                -taxids-only -separate-cols
    read_header | rank | taxon_name              -separate-cols
    read_header | rank | taxon_name | taxon_id   -taxids -separate-cols

    Note that the separator '\t|\t' can be changed to something else with
    the command line option '-separator <text>'.

    Note that the default lowest taxon rank is 'sequence'. Sequence-level taxon
    ids have negative numbers in order to not interfere with NCBI taxon ids.
    Each reference sequence is added as its own taxon below the
    lowest known NCBI taxon for that sequence. If you do not want to classify
    at sequence-level, you can set a higher rank as lowest classification rank
    with the '-lowest' command line option: '-lowest species' or
    '-lowest subspecies' or '-lowest genus', etc.