POPulation Partitioning Using Nucleotide Kmers

<!-- badges: end -->

Description

Links: - Documentation - Databases - Paper

If you find PopPUNK useful, please cite us:

Lees JA, Harris SR, Tonkin-Hill G, Gladstone RA, Lo SW, Weiser JN, Corander J, Bentley SD, Croucher NJ. Fast and flexible bacterial genomic epidemiology with PopPUNK. Genome Research 29:304-316 (2019). doi:10.1101/gr.241455.118

You can also run your command with --citation to get a list of citations and a suggested methods paragraph.

News and roadmap

The roadmap can be found in the documentation.

2024-08-07

PopPUNK 2.7.0 comes with two changes: - Distance matrices <db_name>.dists.npy are no longer required or written when using poppunk_assign, with or without --update-db. These can be very large, especially with many samples, so this saves space and memory in model reuse and distribution. Note that the <db_name>.dists.pkl file is still required (but this is small). - We have added a --stable flag to poppunk_assign. Rather than merging hybrid clusters, new samples will simply be assigned to their nearest neighbour. This implies --serial and cannot be run with --update-db. This behaviour mimics the 'stable nomenclature' of schemes such as LIN.

2023-01-18

We have retired the PopPUNK website. Databases have been expanded, and can be found here: https://www.bacpop.org/poppunk-databases/.

2022-08-04

The change in scikit-learn's API in v1.0.0 and above mean that HDBSCAN models fitted with sklearn <=v0.24 will give an error when loaded. If you run into this, the solution is one of: - Downgrade sklearn to v0.24. - Run model refinement to turn your model into a boundary model instead (this will change clusters). - Refit your model in an environment with sklearn >=v1.0.

If this is a common problem let us know, as we could write a script to 'upgrade' HDBSCAN models. See issue #213 for more details.

2021-03-15

We have fixed a number of bugs with may affect the use of poppunk_assign with --update-db. We have also fixed a number of bugs with GPU distances. These are 'advanced' features and are not likely to be encountered in most cases, but if you do wish to use either of these features please make sure that you are using PopPUNK >=v2.4.0 with pp-sketchlib >=v1.7.0.

2020-09-30

We have discovered a bug affecting the interaction of pp-sketchlib and PopPUNK. If you have used PopPUNK >=v2.0.0 with pp-sketchlib <v1.5.1 label order may be incorrect (see issue #95).

Please upgrade to PopPUNK >=v2.2 and pp-sketchlib >=v1.5.1. If this is not possible, you can either: - Run scripts/poppunk_pickle_fix.py on your .dists.pkl file and re-run model fits. - Create the database with poppunk_sketch directly, rather than PopPUNK --create-db

Installation

This is for the command line version. For more details see installation in the documentation.

Our (beta) web interface BeeBOP is now also available: https://beebop.dide.ic.ac.uk/

Through conda (recommended)

The easiest way is through conda, which is most easily accessed by first installing miniconda. PopPUNK can then be installed by running: conda install poppunk If the package cannot be found you will need to add the necessary channels: conda config --add channels defaults conda config --add channels bioconda conda config --add channels conda-forge

Quick usage

See the overview first. There are two ways of running:

With a supported species

1) Download an existing database. 2) Run assignment.

With a new species.

1) Create sketches of input. 2) Run QC. 3) Build a model.

Docker image

A docker image is available

docker pull mrcide/poppunk:bacpop-20