Additional data for different isolates and antibiotics

Question

[araina]

Hi,

I am using Kover as a part of a course project. I got the tutorial up and running for the sample dataset provided. I just wanted to know where can we find additional data to feed into the model, specifically in the format of KmerMatrix.tsv and metadat.tsv? Are there any sources that you could possible suggest? I am looking to train the model on more bacteria and antibiotics.

Link to Kover - http://aldro61.github.io/kover/doc_example.html

Thanks

[GenoMax]

It is good to include links for a program if it is not something commonly used (as in this case).

[araina]

I have edited the question to include the link for the same.

Answer 1

[Alexandre Drouin]

Hi!

Yes, you can definitely get more data to train kover. The best source for antibiotic resistance datasets is the PATRIC database: https://www.patricbrc.org. I created a Python package to extract data from it: https://github.com/aldro61/patric_tools.

Particularly, have a look at this script that downloads a bunch of genomes with antibiotic resistance labels: https://github.com/aldro61/patric_tools/blob/master/examples/download_amr_datasets.py. You can find the results that I obtained with Kover here: https://aldro61.github.io/kover-amr-platform/. I regularly update them when new data is available. See also this paper that explains the method and the web tool: https://arxiv.org/abs/1612.01030.

The genomes that are downloaded by this script are in FASTA format (a set of sequences). You can train Kover directly from those (see https://aldro61.github.io/kover/doc_dataset.html#from-contigs ). If you want to convert the FASTA files into a big matrix (like KmerMatrix.tsv), you need to use Ray Surveyor (https://github.com/zorino/raysurveyor-tutorial ).

I hope that this helps!

Cheers, Alex

[araina]

Hi Alex, thank you so much for getting back. This definitely helps a lot!