Pathogen-of-origin prediction

In this section we will focus on pathogen-of-origin prediction of AMR genes using q2-rgi.

CARD’s WildCARD Data provide a data set of AMR alleles and their distribution among pathogens and plasmids. CARD’s k-mer classifiers sub-sample these sequences to identify k-mers that are uniquely found within AMR alleles of individual pathogen species, pathogen genera, pathogen-restricted plasmids, or promiscuous plasmids. CARD’s k-mer classifiers can then be used to predict the pathogen of origin for matches found by RGI for MAGs or reads. The k-mer classifiers are available as precomputed databases for 15-mers and 61-mers. Pathogen classification accuracy plateaus at under 80% above a k-mer size of 15, while the k-mer size of 15 represents a reliable minimum k-mer size. Classification accuracy increases with increasing k-mer size, but so does the run time and memory usage. For more information please refer to the RGI documentation and the CARD k-mer paper. For this tutorial we use the 15-mer database because it is faster and requires less memory.

Metagenomic reads¶

To analyse AMR annotations from metagenomic reads we can use the action kmer-query-reads-card.

qiime rgi kmer-query-reads-card \
    --i-amr-annotations rgi_allele_annotations_reads.qza \
    --i-card-db card_db.qza \
    --i-kmer-db 15mer_db.qza \
    --o-reads-allele-kmer-analysis reads_allele_kmer_analysis.qza \
    --o-reads-gene-kmer-analysis reads_gene_kmer_analysis.qza \
    --p-threads 2 \
    --parallel-config parallel.config.toml \
    --verbose

qiime rgi kmer-query-reads-card \
    --i-amr-annotations rgi_allele_annotations_reads.qza \
    --i-card-db card_db.qza \
    --i-kmer-db 15mer_db.qza \
    --o-reads-allele-kmer-analysis reads_allele_kmer_analysis.qza \
    --o-reads-gene-kmer-analysis reads_gene_kmer_analysis.qza \
    --p-threads 2 \
    --verbose

Tabulate results¶

Now we can tabulate the results with the tabulate visualizer.

qiime metadata tabulate \
  --m-input-file reads_gene_kmer_analysis.qza \
  --o-visualization reads_gene_kmer_analysis_tabulated.qzv

Your visualization should look similar to this one.

MAGs¶

To analyse AMR annotations from MAGs we can use the action kmer-query-mags-card.

qiime rgi kmer-query-mags-card \
    --i-amr-annotations rgi_annotations_mags.qza \
    --i-card-db card_db.qza \
    --i-kmer-db 15mer_db.qza \
    --o-mags-kmer-analysis mags_kmer_analysis.qza \
    --p-threads 2 \
    --parallel-config parallel.config.toml \
    --verbose

qiime rgi kmer-query-mags-card \
    --i-amr-annotations rgi_annotations_mags.qza \
    --i-card-db card_db.qza \
    --i-kmer-db 15mer_db.qza \
    --o-mags-kmer-analysis mags_kmer_analysis.qza \
    --p-threads 2 \
    --verbose

Tabulate results¶

Now we can again tabulate the results with the tabulate visualizer.

qiime metadata tabulate \
  --m-input-file mags_kmer_analysis.qza \
  --o-visualization mags_kmer_analysis_tabulated.qzv

Your visualization should look similar to this one.

Build a custom k-mer database¶

While CARD’s WildCARD data set provides a precompiled 15-mer and 61-mer database it is also possible to build custom k-mer databases. This action can be computationally expensive and take a long time to complete, depending on the chosen k-mer size. Because of this it is not feasible to run this action on a personal computer. To see benchmarks released by the developers of RGI please check the CARD k-mer paper. The action kmer-build-card can be used to build a custom k-mer database.

qiime rgi kmer-build-card \
    --i-card-db card_db.qza \
    --p-kmer-size 80 \
    --o-kmer-db 80mer_db.qza \
    --p-threads 40 \
    --verbose