pykraken2

pykraken2 provides a server/client implementation of kraken2.

Installation

pykraken2 is pre-alpha software, there are currently no packaged versions. Please use the Development instructions before for installation and use.

Development

The Git repository contains Python code implementing a server and client pair, it also contains a fork of kraken2 as a git submodule. To obtain the repository:

git clone --recursive https://github.com/epi2me-labs/pykraken2.git

For the purposes of development the Python components of pykraken2 can be installed using an in-place (editable) install:

make develop

This will make a virtual environment at ./venv and create and inplace (editable) install of the Python code, along with compiling kraken2 and copying the executables to the bin directory of the virtual environment.

Usage

Two command-line entry points are provided:

pykraken2 server

to run a kraken2 service to classify reads, and

pykraken2 client

to send read data to the service and receive results.

API

The server and client are straight-forward, we recommend the context managers are used:

# create a server
from pykraken2.server import Server
with Server(database, address, port, threads=8) as server:
    while True:
        pass

Although this code looks a little goofy, we intend to change the behaviour of the context manager. database is a database directory created by kraken2, address a location on which to listen for requests (e.g. 'localhost), port the network port on which to listen, and threads the number of threads to kraken2 to use. An optional argument k2_binary can be used to specify the location of a kraken2 binary to use.

A client can be constructed as:

# create a client
from pykraken2.client import Client
with Client(address, ports) as client:
    with open('output.txt', 'w') as fh:
        for chunk in client.process_fastq(args.fastq):
            fh.write(chunk)

Currently the process_fastq iterator returns chunks of the kraken2 output. The chunks are verbatim output: a chunk may contain multiple records, the last record may be incomplete in any non-final chunk.