Database

class pyskani.Database(path=None, *, compression=125, marker_compression=1000, k=15, format=None)

A database storing sketched genomes.

The database contains two different sketch collections with different compression levels: marker sketches, which are heavily compressed, and always kept in memory; and genome sketches, which take more memory, but may be stored inside an external file.

flush()

Flush the database buffers to disk.

This does nothing for a database loaded in memory. For a database stored in a folder, this will save the markers into a file named markers.bin.

load()

Load a database from a folder containing sketches.

The sketches will be loaded in memory to speed-up querying. To reduce memory consumption and load sketches lazily from the folder, use Database.open.

Parameters:

path (str, bytes, or os.PathLike) – The path to the folder containing the sketched references.

Returns:

Database – A database with all sketches loaded in memory.

Raises:

• OSError – When the files from the folder could not be opened.

• ValueError – When the sketches could not be deserialized.

open()

Open a database from a folder containing sketches.

The marker sketches will be loaded in memory, but the sketches will be loaded only when needed when querying. To speed-up querying by pre-fetching sketches, use Database.load.

Parameters:

path (str, bytes, or os.PathLike) – The path to the folder containing the sketched references.

Returns:

Database – A database with only markers loaded in memory.

Raises:

• OSError – When the files from the folder could not be opened.

• ValueError – When the markers could not be deserialized.

query(name, *contigs, seed=True, learned_ani=None, median=False, robust=False, cutoff=None, faster_small=False)

Query the database with a genome.

Parameters:

• name (str) – The name of the query genome.

• contigs (str, bytes, bytearray or memoryview) – The contigs of the query genome.

Keyword Arguments:

• seed (bool) – Compute seed positions while sketching the query.

• learned_ani (bool or None) – Use a regression model to compute ANI, using a model trained on MAGs. Pass True or False to force enabling or disabling the model, respectively. By default, the regression model is enabled when the sketch compression factor is >=70 and not running in median mode.

• median (bool) – Estimate median identity instead of average identity. Disabled by default. Equivalent to the --median flag of the CLI.

• robust (bool) – Estimate mean after trim off 10%/90% quantiles. Disabled by default. Equivalent to the --robust flag of the CLI.

• cutoff (float or None) – The cutoff to use to screen out pairs with approximately lower identity, as computed with k-mer sketching. Defaults to 0.8 for ANI and 0.6 for AAI. Equivalent to the -s flag from the CLI.

• faster_small (bool) – Set to True to filter genomes with less than 20 marker k-mers more aggressively. Disabled by default. Equivalent to the --faster-small flag of the CLI.

Returns:

list of Hit – The hits found for the query.

..versionadded:: 0.2.0

The cutoff and faster_small keyword arguments.

save(path, overwrite=False, format=None)

Save the database to the given path.

sketch(name, *contigs, seed=True)

Add a reference genome to the database.

Parameters:

• name (str) – The name of the reference genome to add.

• contigs (str, bytes, bytearray or memoryview) – The contigs of the reference genome.

Keyword Arguments:

seed (bool) – Compute seed positions while sketching the query.

compression

The compression factor for sketches.

Type:

int

marker_compression

The compression factor for marker k-mers.

Type:

int

path

The path where sketches are stored.

Type:

pathlib.Path or None