Offline curation tools

Automated offline sequence tagging

Sequence tagging is the process of identifying alleles by scanning the sequence bin linked to an isolate record. Loci need to be defined in an external sequence definition database that contains the sequences for known alleles. The tagging function uses BLAST to identify sequences and will tag the specific sequence region with locus information and an allele designation if a matching allele is identified by reference to an external database.

There is a script called ‘autotag.pl’ in the BIGSdb package. This can be used to tag genome sequences from the command line.

Before autotag.pl can be run for the first time, a log file needs to be created. This can be created if it doesn’t already exist with the following:

sudo touch /var/log/bigsdb_scripts.log
sudo chown bigsdb /var/log/bigsdb_scripts.log

The autotag.pl script should be installed in /usr/local/bin. It is run as follows:

autotag.pl --database <database configuration>

where <database configuration> is the name used for the argument ‘db’ when using the BIGSdb application.

If you have multiple processor cores available, use the –threads option to set the number of jobs to run in parallel. Isolates for scanning will be split among the threads.

The script must be run by a user that can both write to the log file and access the databases, e.g. the ‘bigsdb’ user (see ‘Setting up the offline job manager’).

A full list of options can be found by typing:

autotag.pl --help

NAME
    autotag.pl - BIGSdb automated allele tagger

SYNOPSIS
    autotag.pl --database NAME [options]

OPTIONS
-0, --missing
    Marks missing loci as provisional allele 0. Sets default word size to 15.

-d, --database NAME
    Database configuration name.

-e, --exemplar
    Only use alleles with the 'exemplar' flag set in BLAST searches to identify
    locus within genome. Specific allele is then identified using a database
    lookup. This may be quicker than using all alleles for the BLAST search,
    but will be at the expense of sensitivity. If no exemplar alleles are set
    for a locus then all alleles will be used. Sets default word size to 15.

-f --fast
    Perform single BLAST query against all selected loci together. This will
    take longer to return any results but the overall scan should finish
    quicker. This method will also use more memory - this can be used with
    --exemplar to mitigate against this.

-h, --help
    This help page.

-i, --isolates LIST
    Comma-separated list of isolate ids to scan (ignored if -p used).

--isolate_list_file FILE
    File containing list of isolate ids (ignored if -i or -p used).

-I, --exclude_isolates LIST
    Comma-separated list of isolate ids to ignore.

-l, --loci LIST
    Comma-separated list of loci to scan (ignored if -s used).

-L, --exclude_loci LIST
    Comma-separated list of loci to exclude

-m, --min_size SIZE
    Minimum size of seqbin (bp) - limit search to isolates with at least this
    much sequence.

-n, --new_only
    New (previously untagged) isolates only.  Combine with --new_max_alleles
    if required.

--new_max_alleles ALLELES
    Set the maximum number of alleles that can be designated or sequences
    tagged before an isolate is not considered new when using the --new_only
    option.

-o, --order
    Order so that isolates last tagged the longest time ago get scanned first
    (ignored if -r used).

--only_already_tagged
    Only check loci that already have a tag present (but no allele designation).
    This must be combined with the --already_tagged option or no loci will
    match. This option is used to perform a catch-up scan where a curator has
    previously tagged sequence regions prior to alleles being defined, without
    the need to scan all missing loci.

-p, --projects LIST
    Comma-separated list of project isolates to scan.

-P, --exclude_projects LIST
    Comma-separated list of projects whose isolates will be excluded.

-q, --quiet
    Only error messages displayed.

-r, --random
    Shuffle order of isolate ids to scan.

--reuse_blast
    Reuse the BLAST database for every isolate (when running --fast option).
    All loci will be scanned rather than just those missing from an isolate.
    Consequently, this may be slower if isolates have already been scanned,
    and for the first isolate scanned by a thread. On larger schemes, such as
    wgMLST, or when isolates have not been previously scanned, setting up the
    BLAST database can take a significant amount of time, so this may be
    quicker. This option is always selected if --new_only is used.

-R, --locus_regex REGEX
    Regex for locus names.

-s, --schemes LIST
    Comma-separated list of scheme loci to scan.

-t, --time MINS
    Stop after t minutes.

--threads THREADS
    Maximum number of threads to use.

-T, --already_tagged
    Scan even when sequence tagged (no designation).

-v, --view VIEW
    Isolate database view (overrides value set in config.xml).

-w, --word_size SIZE
    BLASTN word size.

-x, --min ID
    Minimum isolate id.

-y, --max ID
    Maximum isolate id.

Defining exemplar alleles

Exemplar alleles are a subset of the total number of alleles defined for a locus that encompass the known diversity within a specified identity threshold. They can be used to speed up autotagging as the BLAST queries are performed against exemplars to identify the locus region in the genome followed by a direct database lookup of the sequence found to identify the exact allele found. This is usually combined with the autotagger –fast option.

Once exemplars have been defined you may also wish to set the fast_scan=”yes” option in the config.xml file. This enables their use for scanning within the web curators’ interface.

There is a script called ‘find_exemplars.pl’ in the BIGSdb scripts/maintenance directory.

A full list of options can be found by typing:

find_exemplars.pl --help

NAME
    find_exemplars.pl - Identify and mark exemplar alleles for use
    by tagging functions

SYNOPSIS
    find_exemplars.pl --database NAME   [options]

OPTIONS

--database NAME
    Database configuration name.

--datatype DNA|peptide
    Only define exemplars for specified data type (DNA or peptide)

--exclude_loci LIST
    Comma-separated list of loci to exclude

--help
    This help page.

--loci LIST
    Comma-separated list of loci to scan (ignored if -s used).

--locus_regex REGEX
    Regex for locus names.

--schemes LIST
    Comma-separated list of scheme loci to scan.

--update
    Update exemplar flags in database.

--variation DISSIMILARITY
    Value for percentage identity variation that exemplar alleles
    cover (smaller value will result in more exemplars). Default: 10.

Automated offline allele definition

There is a script called ‘scannew.pl’ in the BIGSdb scripts/automation directory. This can be used to identify new alleles from the command line. This can (optionally) upload these to a sequence definition database.

Before scannew.pl can be run for the first time, a log file needs to be created. This can be created if it doesn’t already exist with the following:

sudo touch /var/log/bigsdb_scripts.log
sudo chown bigsdb /var/log/bigsdb_scripts.log

The scannew.pl script should be installed in /usr/local/bin. It is run as follows:

scannew.pl --database <database configuration>

where <database configuration> is the name used for the argument ‘db’ when using the BIGSdb application.

If you have multiple processor cores available, use the –threads option to set the number of jobs to run in parallel. Loci for scanning will be split among the threads.

The script must be run by a user that can both write to the log file and access the databases, e.g. the ‘bigsdb’ user (see ‘Setting up the offline job manager’).

A full list of options can be found by typing:

scannew.pl --help

NAME
  scannew.pl - BIGSdb automated allele definer

SYNOPSIS
  scannew.pl --database NAME [options]

OPTIONS
-a, --assign
    Assign new alleles in definitions database.

--allow_frameshift
    Allow sequences to contain a frameshift so that the length is not a
    multiple of 3, or an internal stop codon. To be used with
    --coding_sequences option to allow automated curation of pseudogenes.
    New alleles assigned will be flagged either 'frameshift' or 'internal stop
    codon' if appropriate.  Essentially, combining these two options only
    checks that the sequence starts with a start codon and ends with a stop
    codon.

--allow_subsequences
    Allow definition of sub- or super-sequences. By default these will not
    be assigned.

-A, --alignment INT
    Percentage alignment (default: 100).

-B, --identity INT
    Percentage identity (default: 99).

-c, --coding_sequences
    Only return complete coding sequences.

-d, --database NAME
    Database configuration name.

-h, --help
    This help page.

-i, --isolates LIST
    Comma-separated list of isolate ids to scan (ignored if -p used).

--isolate_list_file FILE
    File containing list of isolate ids (ignored if -i or -p used).

-I, --exclude_isolates LIST
    Comma-separated list of isolate ids to ignore.

-l, --loci LIST
    Comma-separated list of loci to scan (ignored if -s used).

-L, --exclude_loci LIST
    Comma-separated list of loci to exclude.

-m, --min_size SIZE
    Minimum size of seqbin (bp) - limit search to isolates with at least this
    much sequence.

-n, --new_only
    New (previously untagged) isolates only.

-o, --order
    Order so that isolates last tagged the longest time ago get scanned first
    (ignored if -r used).

-p, --projects LIST
    Comma-separated list of project isolates to scan.

-P, --exclude_projects LIST
    Comma-separated list of projects whose isolates will be excluded.

-q, --quiet
    Only error messages displayed.

-r, --random
    Shuffle order of isolate ids to scan.

-R, --locus_regex REGEX
    Regex for locus names.

-s, --schemes LIST
    Comma-separated list of scheme loci to scan.

-t, --time MINS
    Stop after t minutes.

--threads THREADS
    Maximum number of threads to use.

--type_alleles
    Only use alleles with the 'type_allele' flag set to identify locus.
    If a partial match is found then a full database lookup will be performed
    to identify any known alleles. Using this option will constrain the search
    space so that allele definitions don't become more variable over time. Note
    that you must have at least one allele defined as a type allele for a locus
    if you use this option otherwise you will not find any matches!

-T, --already_tagged
    Scan even when sequence tagged (no designation).

-v, --view VIEW
    Isolate database view (overrides value set in config.xml).

-w, --word_size SIZE
    BLASTN word size.

-x, --min ID
    Minimum isolate id.

-y, --max ID
    Maximum isolate id.

Calculating assembly stats

Basic assembly statistics are calculated automatically by the database engine as contigs are added to the sequence bin. These include the number of contigs, total length and the N50 value. Some calculations, such as %GC, number of Ns, and the number of gaps however, require offline analysis since these involve inspecting the nucleotide content of each contig. These can be calculated by running the update_assembly_stats.pl script. You can choose to run this against all databases on the system or against a specific database.

Only one copy of the script can run at a time, but it will stop gracefully if it detects another copy running, so it is recommended that the script is run regularly using a CRON job and the –quiet option. This ensures that records are updated shortly after they have been uploaded.

Once calculated, all assembly statistics can then be used in isolate queries.

A full list of options can be found by typing:

update_assembly_stats.pl --help

NAME
    update_assembly_stats.pl - Perform/update calculation of
    assembly GC, N and gap stats.

SYNOPSIS
    update_assembly_stats.pl [options]

OPTIONS

--database DATABASE CONFIG
    Database configuration name. If not included then all isolate databases
    defined on the system will be checked.

--exclude CONFIG NAMES
    Comma-separated list of config names to exclude.

--help
    This help page.

--quiet
    Only show errors.

--refresh_days DAYS
    Refresh records last analysed longer that the number of days set. By
    default, only records that have not been analysed will be checked.

Predicting species based on rMLST analysis

The rMLST plugin predicts species based on matches to rMLST alleles exclusively found in a particular species. It uses the PubMLST API to query either a genome sequence or rMLST allele designations to identify the species. When the analysis is run using the plugin, the results are also stored with the isolate record and can then be displayed within the isolate information page. This analysis can also be run offline using the update_rmlst_species.pl script.

Only one copy of the script can run at a time, but it will stop gracefully if it detects another copy running, so it is recommended that the script is run regularly using a CRON job and the –quiet option. This ensures that records are updated shortly after they have been uploaded.

A full list of options can be found by typing:

update_rmlst_species.pl --help

NAME
    update_rmlst_species.pl - Perform/update species id check

SYNOPSIS
    update_rmlst_species.pl [options]

OPTIONS

--database DATABASE CONFIG
    Database configuration name. If not included then all isolate databases
    defined on the system will be checked.

--exclude CONFIG NAMES
    Comma-separated list of config names to exclude.

--help
    This help page.

--last_run_days DAYS
    Only run for a particular isolate when the analysis was last performed
    at least the specified number of days ago.

--quiet
    Only show errors.

--refresh_days DAYS
    Refresh records last analysed longer that the number of days set. By
    default, only records that have not been analysed will be checked.

Cleanly interrupting offline curation

Sometimes you may wish to stop running autotagger or allele autodefiner jobs as they can be run for a long time and as CRON jobs. If these are running in single threaded mode, the easiest way is to simply send a kill signal to the process, i.e. identify the process id using ‘top’, e.g. 23232 and then

kill 23232

The scripts should respond to this signal within a couple of seconds, clean up all their temporary files and write the history log (where appropriate). Do not use ‘kill -9’ as this will terminate the processes immediately and not allow them to clean up.

If these scripts are running using multiple threads, then you need to cleanly kill each of these. The simplest way to terminate all autotagger jobs is to type

pkill autotag

The parent process will wait for all forked processes to cleanly terminate and then exit itself.

Similarly, to terminate all allele autodefiner jobs, type

pkill scannew

Uploading contigs from the command line

There is a script called upload_contigs.pl in the BIGSdb scripts/maintenance directory. This can be used to upload contigs from a local FASTA file for a specified isolate record.

The upload_contigs.pl script should be installed in /usr/local/bin. It is run as follows:

upload_contigs.pl --database <NAME> --isolate <ID> --file <FILE>
    --curator <ID> --sender <ID>

The script must be run by a user who has the appropriate database permissions and the local configuration settings should be modified to match the database user account to be used. The default setting uses the ‘apache’ user which is used by the BIGSdb web interface.

A full list of options can be found by typing:

upload_contigs.pl --help

NAME
    upload_contigs.pl - Upload contigs to BIGSdb isolate database

SYNOPSIS
    upload_contigs.pl --database NAME --isolate ID --file FILE
         --curator ID --sender ID [options]

OPTIONS
-a, --append
    Upload contigs even if isolate already has sequences in the bin.

-c, --curator ID
    Curator id number.

-d, --database NAME
    Database configuration name.

-f, --file FILE
    Full path and filename of contig file.

-h, --help
    This help page.

-i, --isolate ID
    Isolate id of record to upload to.

-m, --method METHOD
    Method, e.g. 'Illumina', default 'unknown'.

--min_length LENGTH
    Exclude contigs with length less than value.

-s, --sender ID
    Sender id number.

Populating geographic point lookup values

If a field has the geographic_point_lookup attribute set to ‘yes’ in the config.xml file, field values can be mapped to GPS coordinates to facilitate mapping.

These lookup values can be populated using the gp_town_lookups.pl script found in the scripts/maintenance directory. This script uses the Geonames dataset that contains GPS coordinates for towns and cities with populations of at least 1000 people. The dataset is included in the datasets/Geonames directory.

A full list of options can be found by typing:

gp_town_lookups.pl --help

NAME
    geography_point_lookups.pl - Populate geography_point_lookup table
    to set city/town GPS coordinates for mapping.

SYNOPSIS
    geography_point_lookups.pl --database NAME --field FIELD --geodataset DIR

    Run this to populate any unassigned values in the geography_point_lookup
    table.

OPTIONS

--database NAME
    Database configuration name.

--feature_code CODE
    Geonames feature code. See http://www.geonames.org/export/codes.html.
    Default is 'P' (towns/cities).

--field FIELD
    Name of field. This should have the geography_point_lookup attribute set to
    'yes' in config.xml.

--geodataset DIR
    Directory containing the Geonames dataset.

--help
    This help page.

--min_population POPULATION
    Set the minimum population for town to assign. Note that all entries in the
    Geonames database has population, so setting this attribute may result in
    some values not being assigned, but can ensure that only high-confidence
    values are used.

--quiet
    Only show error messages.

--tmp_dir DIR
    Location for temporary files. Defaults to /var/tmp/.