Using sourmash from the command line

:depth: 3

From the command line, sourmash can be used to compute MinHash sketches from DNA sequences, compare them to each other, and plot the results; these sketches are saved into "signature files". These signatures allow you to estimate sequence similarity quickly and accurately in large collections, among other capabilities.

Please see the mash software and the mash paper (Ondov et al., 2016) for background information on how and why MinHash sketches work.

sourmash uses a subcommand syntax, so all commands start with sourmash followed by a subcommand specifying the action to be taken.

An example

Grab three bacterial genomes from NCBI:

curl -L -O ftp://ftp.ncbi.nlm.nih.gov/genomes/refseq/bacteria/Escherichia_coli/reference/GCF_000005845.2_ASM584v2/GCF_000005845.2_ASM584v2_genomic.fna.gz
curl -L -O ftp://ftp.ncbi.nlm.nih.gov/genomes/refseq/bacteria/Salmonella_enterica/reference/GCF_000006945.2_ASM694v2/GCF_000006945.2_ASM694v2_genomic.fna.gz
curl -L -O ftp://ftp.ncbi.nlm.nih.gov/genomes/refseq/bacteria/Sphingobacteriaceae_bacterium_DW12/latest_assembly_versions/GCF_000783305.1_ASM78330v1/GCF_000783305.1_ASM78330v1_genomic.fna.gz

Compute signatures for each:

   sourmash compute -k 31 *.fna.gz

This will produce three .sig files containing MinHash signatures at k=31.

Next, compare all the signatures to each other:

sourmash compare *.sig -o cmp

Optionally, parallelize compare to 8 threads with -p 8:

sourmash compare -p 8 *.sig -o cmp

Finally, plot a dendrogram:

sourmash plot cmp --labels

This will output two files, cmp.dendro.png and cmp.matrix.png, containing a clustering & dendrogram of the sequences, as well as a similarity matrix and heatmap.

Matrix:

The `sourmash` command and its subcommands

To get a list of subcommands, run sourmash without any arguments.

There are five main subcommands: compute, compare, plot, search, and gather. See the tutorial for a walkthrough of these commands.

compute creates signatures.
compare compares signatures and builds a distance matrix.
plot plots distance matrices created by compare.
search finds matches to a query signature in a collection of signatures.
gather finds non-overlapping matches to a metagenome in a collection of signatures.

There are also a number of commands that work with taxonomic information; these are grouped under the sourmash lca subcommand. See the LCA tutorial for a walkthrough of these commands.

lca classify classifies many signatures against an LCA database.
lca summarize summarizes the content of metagenomes using an LCA database.
lca gather finds non-overlapping matches to a metagenome in an LCA database.
lca index creates a database for use with LCA subcommands.
lca rankinfo summarizes the content of a database.
lca compare_csv compares lineage spreadsheets, e.g. those output by lca classify.

Finally, there are a number of utility and information commands:

info shows version and software information.
index indexes many signatures using a Sequence Bloom Tree (SBT).
sbt_combine combines multiple SBTs.
categorize is an experimental command to categorize many signatures.
watch is an experimental command to classify a stream of sequencing data.

Please use the command line option --help to get more detailed usage information for each command.

Note that as of sourmash v3.4, most commands will load signatures from indexed databases (the SBT and LCA formats) as well as from signature files.

`sourmash compute` - make sourmash signatures from sequence data

The compute subcommand computes and saves signatures for each sequence in one or more sequence files. It takes as input FASTA or FASTQ files, and these files can be uncompressed or compressed with gzip or bzip2. The output will be one or more JSON signature files that can be used with sourmash compare.

Please see Using sourmash: a practical guide for more information on computing signatures.

Usage:

sourmash compute filename [ filename2 ... ]

Optional arguments:

--ksizes K1[,K2,K3] -- one or more k-mer sizes to use; default is 31
--force -- recompute existing signatures; convert non-DNA characters to N
--output -- save all the signatures to this file; can be '-' for stdout.
--track-abundance -- compute and save k-mer abundances.
--name-from-first -- name the signature based on the first sequence in the file
--singleton -- instead of computing a single signature for each input file,
               compute one for each sequence
--merged <name> -- compute a single signature for all of the input files,
                   naming it <name>

`sourmash compare` - compare many signatures

The compare subcommand compares one or more signatures (created with compute) using estimated Jaccard index or (if signatures are computed with --track-abundance) the angular similarity.

The default output is a text display of a similarity matrix where each entry [i, j] contains the estimated Jaccard index between input signature i and input signature j. The output matrix can be saved to a file with --output and used with the sourmash plot subcommand (or loaded with numpy.load(...). Using --csv will output a CSV file that can be loaded into other languages than Python, such as R.

Usage:

sourmash compare file1.sig [ file2.sig ... ]

Options:

--output -- save the distance matrix to this file (as a numpy binary matrix)
--ksize -- do the comparisons at this k-mer size.
--containment -- compute containment instead of similarity.
        C(i, j) = size(i intersection j) / size(i).
--from-file -- append the list of files in this text file to the input
        signatures

Note: compare by default produces a symmetric similarity matrix that can be used as an input to clustering. With --containment, however, this matrix is no longer symmetric and cannot formally be used for clustering.

`sourmash plot` - cluster and visualize comparisons of many signatures

The plot subcommand produces two plots -- a dendrogram and a dendrogram+matrix -- from a distance matrix computed by sourmash compare --output <matrix>. The default output is two PNG files.

Usage:

sourmash plot <matrix>

Options:

--pdf -- output PDF files.
--labels -- display the signature names (by default, the filenames) on the plot
--indices -- turn off index display on the plot.
--vmax -- maximum value (default 1.0) for heatmap.
--vmin -- minimum value (default 0.0) for heatmap.
--subsample=<N> -- plot a maximum of <N> samples, randomly chosen.
--subsample-seed=<seed> -- seed for pseudorandom number generator.

Example output:

`sourmash search` - search for signatures in collections or databases

The search subcommand searches a collection of signatures or SBTs for matches to the query signature. It can search for matches with either high Jaccard similarity or containment; the default is to use Jaccard similarity, unless --containment is specified. -o/--output will create a CSV file containing the matches.

search will load all of provided signatures into memory, which can be slow and somewhat memory intensive for large collections. You can use sourmash index to create a Sequence Bloom Tree (SBT) that can be quickly searched on disk; this is the same format in which we provide GenBank and other databases.

Usage:

sourmash search query.sig [ list of signatures or SBTs ]

Example output:

49 matches; showing first 20:
similarity   match
----------   -----
 75.4%       NZ_JMGW01000001.1 Escherichia coli 1-176-05_S4_C2 e117605...
 72.2%       NZ_GG774190.1 Escherichia coli MS 196-1 Scfld2538, whole ...
 71.4%       NZ_JMGU01000001.1 Escherichia coli 2-011-08_S3_C2 e201108...
 70.1%       NZ_JHRU01000001.1 Escherichia coli strain 100854 100854_1...
 69.0%       NZ_JH659569.1 Escherichia coli M919 supercont2.1, whole g...
...

`sourmash gather` - find metagenome members

The gather subcommand finds all non-overlapping matches to the query. This is specifically meant for metagenome and genome bin analysis. (See Classifying Signatures for more information on the different approaches that can be used here.)

If the input signature was computed with --track-abundance, output will be abundance weighted (unless --ignore-abundances is specified). -o/--output will create a CSV file containing the matches.

gather, like search, will load all of provided signatures into memory. You can use sourmash index to create a Sequence Bloom Tree (SBT) that can be quickly searched on disk; this is the same format in which we provide GenBank and other databases.

Usage:

sourmash gather query.sig [ list of signatures or SBTs ]

Example output:

overlap     p_query p_match 
---------   ------- --------
1.4 Mbp      11.0%   58.0%      JANA01000001.1 Fusobacterium sp. OBRC...
1.0 Mbp       7.7%   25.9%      CP001957.1 Haloferax volcanii DS2 pla...
0.9 Mbp       7.4%   11.8%      BA000019.2 Nostoc sp. PCC 7120 DNA, c...
0.7 Mbp       5.9%   23.0%      FOVK01000036.1 Proteiniclasticum rumi...
0.7 Mbp       5.3%   17.6%      AE017285.1 Desulfovibrio vulgaris sub...

The command line option --threshold-bp sets the threshold below which matches are no longer reported; by default, this is set to 50kb. see the Appendix in Classifying Signatures for details.

Note:

Use sourmash gather to classify a metagenome against a collection of genomes with no (or incomplete) taxonomic information. Use sourmash lca summarize and sourmash lca gather to classify a metagenome using a collection of genomes with taxonomic information.

`sourmash lca` subcommands for taxonomic classification

These commands use LCA databases (created with lca index, below, or prepared databases such as genbank-k31.lca.json.gz).

`sourmash lca classify` - classify a genome using an LCA database

sourmash lca classify classifies one or more signatures using the given list of LCA DBs. It is meant for classifying metagenome-assembled genome bins (MAGs) and single-cell genomes (SAGs).

Usage:

sourmash lca classify --query query.sig [query2.sig ...] --db <lca db> [<lca db2> ...]

For example, the command

sourmash lca classify --query tests/test-data/63.fa.sig \
    --db podar-ref.lca.json

will produce the following logging to stderr:

loaded 1 LCA databases. ksize=31, scaled=10000
finding query signatures...
outputting classifications to stdout
... classifying NC_011663.1 Shewanella baltica OS223, complete genome
classified 1 signatures total

and the example classification output is a CSV file with headers:

ID,status,superkingdom,phylum,class,order,family,genus,species
"NC_009665.1 Shewanella baltica OS185, complete genome",found,Bacteria,Proteobacteria,Gammaproteobacteria,Alteromonadales,Shewanellaceae,Shewanella,Shewanella baltica

The status column in the classification output can take three possible values: nomatch, found, and disagree. nomatch means that no match was found for this query, and found means that an unambiguous assignment was found - all k-mers were classified within the same taxonomic hierarchy, and the most detailed lineage available was reported. disagree means that there was a taxonomic disagreement, and the lowest compatible taxonomic node was reported.

To elaborate on this a bit, suppose that all of the k-mers within a signature were classified as family Shewanellaceae, genus Shewanella, or species Shewanella baltica. Then the lowest compatible node (here species Shewanella baltica) would be reported, and the status of the classification would be found. However, if a number of additional k-mers in the input signature were classified as Shewanella oneidensis, sourmash would be unable to resolve the taxonomic assignment below genus Shewanella and it would report a status of disagree with the genus-level assignment of Shewanella; species level assignments would not be reported.

(This is the approach that Kraken and other lowest common ancestor implementations use, we believe.)

Note: you can specify a list of files to load signatures from in a text file passed to sourmash lca classify with the --query-from-file flag; these files will be appended to the --query input.

`sourmash lca summarize` - summarize a metagenome's contents using an LCA database

sourmash lca summarize produces a Kraken-style summary of the combined contents of the given query signatures. It is meant for exploring metagenomes and metagenome-assembled genome bins.

Note, unlike sourmash lca classify, lca summarize merges all of the query signatures into one and reports on the combined contents. To report on individual signatures, use the --singleton flag; this will become default in sourmash 4.0 and beyond, and the merging behavior will be removed.

As of sourmash 3.4, sourmash lca summarize also supports abundance weighted queries; this can be turned on with --with-abundance. This flag will weight output percentages by the number of times a k-mer is seen. This will become default behavior in sourmash 4.0 and beyond.

Usage:

sourmash lca summarize --query query.sig [query2.sig ...] 
    --db <lca db> [<lca db2> ...]

For example, with the data in tests/test-data/fake-abund, the command line:

sourmash lca summarize --query query.sig.gz --db matches.lca.json.gz

will produce the following log output to stderr:

loaded 1 LCA databases. ksize=31, scaled=10000
finding query signatures...
loaded 1 signatures from 1 files total.

and the following example summarize output to stdout:

79.6%   550   Bacteria;Proteobacteria;Gammaproteobacteria;Alteromonadales;Shewanellaceae;Shewanella;Shewanella baltica;Shewanella baltica OS223
79.6%   550   Bacteria;Proteobacteria;Gammaproteobacteria;Alteromonadales;Shewanellaceae;Shewanella;Shewanella baltica
79.6%   550   Bacteria;Proteobacteria;Gammaproteobacteria;Alteromonadales;Shewanellaceae;Shewanella
79.6%   550   Bacteria;Proteobacteria;Gammaproteobacteria;Alteromonadales;Shewanellaceae
79.6%   550   Bacteria;Proteobacteria;Gammaproteobacteria;Alteromonadales
79.6%   550   Bacteria;Proteobacteria;Gammaproteobacteria
79.6%   550   Bacteria;Proteobacteria
79.6%   550   Bacteria
20.4%   141   Archaea;Euryarchaeota;unassigned;unassigned;unassigned;Aciduliprofundum;Aciduliprofundum boonei;Aciduliprofundum boonei T469
20.4%   141   Archaea;Euryarchaeota;unassigned;unassigned;unassigned;Aciduliprofundum;Aciduliprofundum boonei
20.4%   141   Archaea;Euryarchaeota;unassigned;unassigned;unassigned;Aciduliprofundum
20.4%   141   Archaea;Euryarchaeota;unassigned;unassigned;unassigned
20.4%   141   Archaea;Euryarchaeota;unassigned;unassigned
20.4%   141   Archaea;Euryarchaeota;unassigned
20.4%   141   Archaea;Euryarchaeota
20.4%   141   Archaea

The output is space-separated and consists of three columns: the percentage of total k-mers that have this classification; the number of k-mers that have this classification; and the lineage classification. K-mer classifications are reported hierarchically, so the percentages and totals contain all assignments that are at a lower taxonomic level - e.g. Bacteria, above, contains all the k-mers in Bacteria;Proteobacteria.

The same information is reported in a CSV file if -o/--output is used.

The proportions reflect the query signature construction, where the metagenome contains a 1.5 Mbp Archaeal genome and a 5.4 Mbp Bacterial genome. The Archaeal genome is therefore only ~20% of the distinct k-mers in the metagenome (1.5 Mbp divided by 6.9 Mbp).

If --with-abundance is given, the output changes to reflect the proportions of the query metagenome based on k-mer/read abundances:

56.8%   740   Archaea;Euryarchaeota;unassigned;unassigned;unassigned;Aciduliprofundum;Aciduliprofundum boonei;Aciduliprofundum boonei T469
56.8%   740   Archaea;Euryarchaeota;unassigned;unassigned;unassigned;Aciduliprofundum;Aciduliprofundum boonei
56.8%   740   Archaea;Euryarchaeota;unassigned;unassigned;unassigned;Aciduliprofundum
56.8%   740   Archaea;Euryarchaeota;unassigned;unassigned;unassigned
56.8%   740   Archaea;Euryarchaeota;unassigned;unassigned
56.8%   740   Archaea;Euryarchaeota;unassigned
56.8%   740   Archaea;Euryarchaeota
56.8%   740   Archaea
43.2%   563   Bacteria;Proteobacteria;Gammaproteobacteria;Alteromonadales;Shewanellaceae;Shewanella;Shewanella baltica;Shewanella baltica OS223
43.2%   563   Bacteria;Proteobacteria;Gammaproteobacteria;Alteromonadales;Shewanellaceae;Shewanella;Shewanella baltica
43.2%   563   Bacteria;Proteobacteria;Gammaproteobacteria;Alteromonadales;Shewanellaceae;Shewanella
43.2%   563   Bacteria;Proteobacteria;Gammaproteobacteria;Alteromonadales;Shewanellaceae
43.2%   563   Bacteria;Proteobacteria;Gammaproteobacteria;Alteromonadales
43.2%   563   Bacteria;Proteobacteria;Gammaproteobacteria
43.2%   563   Bacteria;Proteobacteria
43.2%   563   Bacteria

Here, the changed proportions reflect the query signature abundances, where the 1.5 Mbp Archaeal genome is present 5 times, while the 5.4 Mbp Bacterial genome is present only once; when weighted by abundance, the Bacterial genome is only 41.8% of the metagenome content, while the Archaeal genome is 58.1% of the metagenome content.

Note: you can specify a list of files to load signatures from in a text file passed to sourmash lca summarize with the --query-from-file flag; these files will be appended to the --query input.

`sourmash lca gather` - find metagenome taxonomy (DEPRECATED for 4.0)

The sourmash lca gather command finds all non-overlapping matches to the query, similar to the sourmash gather command. This is specifically meant for metagenome and genome bin analysis. (See Classifying Signatures for more information on the different approaches that can be used here.)

If the input signature was computed with --track-abundance, output will be abundance weighted (unless --ignore-abundances is specified). -o/--output will create a CSV file containing the matches.

Usage:

sourmash lca gather query.sig [<lca database> ...]

Example output:

overlap     p_query p_match
---------   ------- --------
1.8 Mbp      14.6%    9.1%      Fusobacterium nucleatum
1.0 Mbp       7.8%   16.3%      Proteiniclasticum ruminis
1.0 Mbp       7.7%   25.9%      Haloferax volcanii
0.9 Mbp       7.4%   11.8%      Nostoc sp. PCC 7120
0.9 Mbp       7.0%    5.8%      Shewanella baltica
0.8 Mbp       6.0%    8.6%      Desulfovibrio vulgaris
0.6 Mbp       4.9%   12.6%      Thermus thermophilus

`sourmash lca index` - build an LCA database

The sourmash lca index command creates an LCA database from a lineage spreadsheet and a collection of signatures. This can be used to create LCA databases from private collections of genomes, and can also be used to create databases for e.g. subsets of GenBank.

See the sourmash lca tutorial and the blog post Why are taxonomic assignments so different for Tara bins? for some use cases.

If you are interested in preparing lineage spreadsheets from GenBank genomes (or building off of NCBI taxonomies more generally), please see the NCBI lineage repository.

You can use --from-file to pass lca index a text file containing a list of files to index.

`sourmash lca rankinfo` - examine an LCA database

The sourmash lca rankinfo command displays k-mer specificity information for one or more LCA databases. See the blog post How specific are k-mers for taxonomic assignment of microbes, anyway? for example output.

`sourmash lca compare_csv` - compare taxonomic spreadsheets

The sourmash lca compare_csv command compares two lineage spreadsheets (such as those output by sourmash lca classify or taken as input by sourmash lca index) and summarizes their agreement/disagreement. Please see the blog post Why are taxonomic assignments so different for Tara bins? for an example use case.

`sourmash signature` subcommands for signature manipulation

These commands manipulate signatures from the command line. Currently supported subcommands are merge, rename, intersect, extract, downsample, subtract, import, export, info, flatten, filter, cat, and split.

The signature commands that combine or otherwise have multiple signatures interacting (merge, intersect, subtract) work only on compatible signatures, where the k-mer size and nucleotide/protein sequences match each other. If working directly with the hash values (e.g. merge, intersect, subtract) then the scaled values must also match; you can use downsample to convert a bunch of samples to the same scaled value.

If there are multiple signatures in a file with different ksizes and/or from nucleotide and protein sequences, you can choose amongst them with -k/--ksize and --dna or --protein, as with other sourmash commands such as search, gather, and compare.

Note, you can use sourmash sig as shorthand for all of these commands.

Most commands will load signatures automatically from indexed databases (SBT and LCA formats) as well as from signature files, and you can load signatures from stdin using - on the command line.

`sourmash signature cat` - concatenate multiple signatures together

Concatenate signature files.

For example,

sourmash signature cat file1.sig file2.sig -o all.sig

will combine all signatures in file1.sig and file2.sig and put them in the file all.sig.

`sourmash signature describe` - display detailed information about signatures

Display signature details.

For example,

sourmash sig describe tests/test-data/47.fa.sig

will display:

signature filename: tests/test-data/47.fa.sig
signature: NC_009665.1 Shewanella baltica OS185, complete genome
source file: 47.fa
md5: 09a08691ce52952152f0e866a59f6261
k=31 molecule=DNA num=0 scaled=1000 seed=42 track_abundance=0
size: 5177
signature license: CC0

`sourmash signature split` - split signatures into individual files

Split each signature in the input file(s) into individual files, with standardized names. Note: unlike the rest of the sourmash sig commands, split can load signatures from LCA and SBT databases as well.

For example,

sourmash signature split tests/test-data/2.fa.sig

will create 3 files,

f372e478.k=21.scaled=1000.DNA.dup=0.2.fa.sig, f3a90d4e.k=31.scaled=1000.DNA.dup=0.2.fa.sig, and 43f3b48e.k=51.scaled=1000.DNA.dup=0.2.fa.sig, representing the three different DNA signatures at different ksizes created from the input file 2.fa.

The format of the names of the output files is standardized and stable for major versions of sourmash: currently, they are period-separated with fields:

md5sum - a unique hash value based on the contents of the signature.
k=<ksize> - k-mer size.
scaled=<scaled> or num=<num> - scaled or num value for MinHash.
<moltype> - the molecule type (DNA, protein, dayhoff, or hp)
dup=<n> - a non-negative integer that prevents duplicate signatures from colliding.
basename - basename of first input file used to create signature; if none provided, or stdin, this is none.

If --outdir is specified, all of the signatures are placed in outdir.

`sourmash signature merge` - merge two or more signatures into one

Merge two (or more) signatures.

For example,

sourmash signature merge file1.sig file2.sig -o merged.sig

will output the union of all the hashes in file1.sig and file2.sig to merged.sig.

All of the signatures passed to merge must either have been computed with --track-abundance, or not. If they have track_abundance on, then the merged signature will have the sum of all abundances across the individual signatures. The --flatten flag will override this behavior and allow merging of mixtures by removing all abundances.

`sourmash signature rename` - rename a signature

Rename the display name for one or more signatures - this is the name output for matches in compare, search, gather, etc.

For example,

sourmash signature rename file1.sig "new name" -o renamed.sig

will place a renamed copy of the hashes in file1.sig in the file renamed.sig. If you provide multiple signatures, all will be renamed to the same name.

`sourmash signature subtract` - subtract other signatures from a signature

Subtract all of the hash values from one signature that are in one or more of the others.

For example,

sourmash signature subtract file1.sig file2.sig file3.sig -o subtracted.sig

will subtract all of the hashes in file2.sig and file3.sig from file1.sig, and save the new signature to subtracted.sig.

To use subtract on signatures calculated with --track-abundance, you must specify --flatten.

`sourmash signature intersect` - intersect two (or more) signatures

Output the intersection of the hash values in multiple signature files.

For example,

sourmash signature intersect file1.sig file2.sig file3.sig -o intersect.sig

will output the intersection of all the hashes in those three files to intersect.sig.

The intersect command flattens all signatures, i.e. the abundances in any signatures will be ignored and the output signature will have track_abundance turned off.

`sourmash signature downsample` - decrease the size of a signature

Downsample one or more signatures.

With downsample, you can --

increase the --scaled value for a signature computed with --scaled, shrinking it in size;
decrease the num value for a traditional num MinHash, shrinking it in size;
try to convert a --scaled signature to a num signature;
try to convert a num signature to a --scaled signature.

For example,

sourmash signature downsample file1.sig file2.sig --scaled 100000 -o downsampled.sig

will output each signature, downsampled to a scaled value of 100000, to downsampled.sig; and

sourmash signature downsample --num 500 scaled_file.sig -o downsampled.sig

will try to convert a scaled MinHash to a num MinHash.

`sourmash signature extract` - extract signatures from a collection

Extract the specified signature(s) from a collection of signatures.

For example,

sourmash signature extract *.sig -k 21 --dna -o extracted.sig

will extract all nucleotide signatures calculated at k=21 from all .sig files in the current directory.

There are currently two other useful selectors for extract: you can specify (part of) an md5sum, as output in the CSVs produced by search and gather; and you can specify (part of) a name.

For example,

sourmash signature extract tests/test-data/*.fa.sig --md5 09a0869

will extract the signature from 47.fa.sig which has an md5sum of 09a08691ce52952152f0e866a59f6261; and

sourmash signature extract tests/test-data/*.fa.sig --name NC_009665

will extract the same signature, which has an accession number of NC_009665.1.

`sourmash signature flatten` - remove abundance information from signatures

Flatten the specified signature(s), removing abundances and setting track_abundance to False.

For example,

sourmash signature flatten *.sig -o flattened.sig

will remove all abundances from all of the .sig files in the current directory.

The flatten command accepts the same selectors as extract.

`sourmash signature filter` - remove hashes based on abundance

Filter the hashes in the specified signature(s) by abundance, by either -m/--min-abundance or -M/--max-abundance or both. Abundance selection is inclusive, so -m 2 -M 5 will select hashes with abundance greater than or equal to 2, and less than or equal to 5.

For example,

sourmash signature -m 2 *.sig

will output new signatures containing only hashes that occur two or more times in each signature.

The filter command accepts the same selectors as extract.

`sourmash signature import` - import signatures from mash.

Import signatures into sourmash format. Currently only supports mash, and can import mash sketches output by mash info -d <filename.msh>.

For example,

sourmash signature import filename.msh.json -o imported.sig

will import the contents of filename.msh.json into imported.sig.

`sourmash signature export` - export signatures to mash.

Export signatures from sourmash format. Currently only supports mash dump format.

For example,

sourmash signature export filename.sig -o filename.sig.msh.json

`sourmash signature overlap` - detailed comparison of two signatures' overlap

Display a detailed comparison of two signatures. This computes the Jaccard similarity (as in sourmash compare or sourmash search) and the Jaccard containment in both directions (as with --containment). It also displays the number of hash values in the union and intersection of the two signatures, as well as the number of disjoint hash values in each signature.

This command has two uses - first, it is helpful for understanding how similarity and containment are calculated, and second, it is useful for analyzing signatures with very small overlaps, where the similarity and/or containment might be very close to zero.

For example,

sourmash signature overlap file1.sig file2.sig

will display the detailed comparison of file1.sig and file2.sig.

Advanced command-line usage

Loading signatures and databases

sourmash uses several different command-line styles.

Briefly,

search and gather both take a single query signature and search multiple signatures or databases. In this case, there has to be a single identifiable query for sourmash to use, and if you're using a database or list of signatures as the source of a query, you'll need to provide a selector (ksize with -k, moltype with --dna etc, or md5sum with --query-md5) that picks out a single signature.
compare takes multiple signatures and can load them from files, directories, and indexed databases (SBT or LCA). It can also take a list of file paths in a text file, using --from-file (see below).
the lca classify and lca summarize commands take multiple signatures with --query, and multiple LCA databases, with --db. sourmash multigather also uses this style. This allows these commands to specify multiple queries and multiple databases without (too much) confusion. These commands will take files containing signature files using --query-from-file (see below).
index and lca index take a few fixed parameters (database name, taxonomy spreadsheet) and then an arbitrary number of other files that contain signatures, including files, directories, and indexed databases. These commands will also take --from-file (see below).

None of these commands currently support searching, comparing, or indexing signatures with multiple ksizes or moltypes at the same time; you need to pick the ksize and moltype to use for your search. Where possible, scaled values will be made compatible.

Storing (and searching) signatures

Backing up a little, there are many ways to store and search signatures.

The simplest is one signature in a single JSON file. You can also put many signatures in a single JSON file, either by building them that way with sourmash compute or by using sourmash sig cat or other commands. Searching or comparing these files involves loading them sequentially and iterating across all of the signatures - which can be slow, especially for many (100s or 1000s) of signatures.

Indexed databases can make searching signatures a lot faster. SBT databases are low memory and disk-intensive databases that allow for fast searches using a tree structure, while LCA databases are higher memory and (after a potentially significant load time) are quite fast.

(LCA databases also permit taxonomic searches using sourmash lca functions.)

The main point is that since all of these databases contain signatures, as of sourmash 3.4, any command that takes more than one signature will also automatically load all of the signatures in the database.

Note that, for now, both SBT and LCA database can only contain one "type" of signature (one ksize, one moltype, etc.) If the database signature type is incompatible with the other signatures, sourmash will complain. In contrast, signature files can contain many different types of signatures, and compatible ones will be discovered automatically.

Passing in lists of files

Various sourmash commands will also take --from-file or --query-from-file, which will take a path to a text file containing a list of file paths. This can be useful for situations where you want to specify thousands of queries, or a subset of signatures produced by some other command.

Loading all signatures under a directory

Note that until 4.0, --traverse-directory may be needed for many commands in order for them to load signatures from a directory hierarchy -- search, gather, index, lca index, and compare, for example. All of the sourmash sig commands support loading from a directory if you provide it on the command line, and this will be the default behavior in sourmash 4.0.

Combining search databases on the command line

All of the commands in sourmash operate in "online" mode, so you can combine multiple databases and signatures on the command line and get the same answer as if you built a single large database from all of them. The only addendum to this rule is that if you have multiple identical matches, the first one to be found will differ depending on the order that the files are passed in on the command line.

This can actually be pretty convenient for speeding up searches - for example, if you're using sourmash gather and you want to find any new results after a database update, you can provide a file containing the previously found matches on the command line before the updated database. Then gather will automatically "find" the previously found matches before anything else, but only if there are no better matches to be found in the updated database. (OK, it's a bit of a niche case, but it's been useful. :)

Using stdin

Most commands will take stdin via the usual UNIX convention, -. Moreover, sourmash compute and the sourmash sig commands will output to stdout. So, for example,

sourmash compute ... -o - | sourmash sig describe - will describe the signatures that were just computed.

(This is a relatively new feature as of 3.4 and our testing may need some work, so please let us know if there's something that doesn't work and we will fix it :).

Files

command-line.md

Latest commit

History

command-line.md

File metadata and controls

Using sourmash from the command line

An example

The sourmash command and its subcommands

sourmash compute - make sourmash signatures from sequence data

sourmash compare - compare many signatures

sourmash plot - cluster and visualize comparisons of many signatures

sourmash search - search for signatures in collections or databases

sourmash gather - find metagenome members

sourmash lca subcommands for taxonomic classification

sourmash lca classify - classify a genome using an LCA database

sourmash lca summarize - summarize a metagenome's contents using an LCA database

sourmash lca gather - find metagenome taxonomy (DEPRECATED for 4.0)

sourmash lca index - build an LCA database

sourmash lca rankinfo - examine an LCA database

sourmash lca compare_csv - compare taxonomic spreadsheets

sourmash signature subcommands for signature manipulation

sourmash signature cat - concatenate multiple signatures together

sourmash signature describe - display detailed information about signatures

sourmash signature split - split signatures into individual files

sourmash signature merge - merge two or more signatures into one

sourmash signature rename - rename a signature

sourmash signature subtract - subtract other signatures from a signature

sourmash signature intersect - intersect two (or more) signatures

sourmash signature downsample - decrease the size of a signature

sourmash signature extract - extract signatures from a collection

sourmash signature flatten - remove abundance information from signatures

sourmash signature filter - remove hashes based on abundance

sourmash signature import - import signatures from mash.

sourmash signature export - export signatures to mash.

sourmash signature overlap - detailed comparison of two signatures' overlap

Advanced command-line usage

Loading signatures and databases

Storing (and searching) signatures

Passing in lists of files

Loading all signatures under a directory

Combining search databases on the command line

Using stdin

The `sourmash` command and its subcommands

`sourmash compute` - make sourmash signatures from sequence data

`sourmash compare` - compare many signatures

`sourmash plot` - cluster and visualize comparisons of many signatures

`sourmash search` - search for signatures in collections or databases

`sourmash gather` - find metagenome members

`sourmash lca` subcommands for taxonomic classification

`sourmash lca classify` - classify a genome using an LCA database

`sourmash lca summarize` - summarize a metagenome's contents using an LCA database

`sourmash lca gather` - find metagenome taxonomy (DEPRECATED for 4.0)

`sourmash lca index` - build an LCA database

`sourmash lca rankinfo` - examine an LCA database

`sourmash lca compare_csv` - compare taxonomic spreadsheets

`sourmash signature` subcommands for signature manipulation

`sourmash signature cat` - concatenate multiple signatures together

`sourmash signature describe` - display detailed information about signatures

`sourmash signature split` - split signatures into individual files

`sourmash signature merge` - merge two or more signatures into one

`sourmash signature rename` - rename a signature

`sourmash signature subtract` - subtract other signatures from a signature

`sourmash signature intersect` - intersect two (or more) signatures

`sourmash signature downsample` - decrease the size of a signature

`sourmash signature extract` - extract signatures from a collection

`sourmash signature flatten` - remove abundance information from signatures

`sourmash signature filter` - remove hashes based on abundance

`sourmash signature import` - import signatures from mash.

`sourmash signature export` - export signatures to mash.

`sourmash signature overlap` - detailed comparison of two signatures' overlap