Contig-based AMR abundance estimation

In this section, we focus on estimating per-sample abundances of AMR features detected in assembled contigs. For this, we will estimate contig abundances from read mappings, normalize them by contig length and link them to AMR annotations to obtain an abundance table of AMR features across samples. This section uses plugins included in the pathogenome and the MOSHPIT distribution of QIIME 2. You can find detailed installation instructions on the QIIME 2 Library.

Build AMR feature table¶

First we convert the AMRFinderPlus annotation results into a feature table that is indexed by contig IDs. You have to run this step with the pathogenome distribution.

qiime amrfinderplus create-feature-table \
    --i-annotations amrfinderplus_annotations.qza \
    --o-table amrfinderplus_annotations_ft.qza \
    --verbose

Get contig lengths¶

Before we can estimate contig abundances, we first need to determine the length of each contig in the assembly. These lengths are used to normalize the abundances. Without this normalisation longer contigs would appear more abundant simply because they can cover more reads.

mosh annotate get-feature-lengths \
    --i-features contigs.qza \
    --o-lengths contigs_lengths.qza \
    --verbose

Estimate contig abundances¶

Next, we estimate how abundant each contig is in every sample. To do this, we map reads back to the contigs and normalize the counts by contig length. This produces a per-sample abundance table with samples as rows and contigs as columns.

mosh annotate estimate-abundance \
    --i-alignment-maps reads-to-contigs-aln.qza \
    --i-feature-lengths contigs_lengths.qza \
    --o-abundances contig_abundances.qza \
    --verbose

Filter contig abundances by AMR annotations¶

At this point, the contig abundance table contains all contigs, but we are only interested in those carrying AMR annotations. Because of this, we have to filter the contig abundance table to only retain contigs with AMR annotations.

mosh feature-table filter-features \
    --i-table contig_abundances.qza \
    --m-metadata-file amrfinderplus_annotations_ft.qza \
    --o-filtered-table contig_abundances_filtered.qza \
    --verbose

Compute per sample AMR abundances¶

Now that we have contig abundances and AMR annotations, we combine both to estimate the abundance of each AMR feature per sample. By multiplying the filtered contig abundance table with the AMR feature table, we obtain a new table in which the rows correspond to samples, the columns to AMR features, and the values to their abundances.

mosh annotate multiply-tables \
    --i-table1 contig_abundances_filtered.qza \
    --i-table2 amrfinderplus_annotations_ft.qza \
    --o-result-table amrfinderplus_abundances_ft.qza \
    --verbose

Create Heatmap¶

To visualize these results, we can generate a heatmap that displays AMR gene abundances across samples. For better visualization, the abundances are log transformed.

mosh feature-table heatmap \
    --i-table amrfinderplus_abundances_ft.qza \
    --o-visualization amrfinderplus_heatmap.qzv

Your visualization should look similar to this one.

Summarize abundances¶

Lastly, we can generate a summarization of the AMR gene abundances.

mosh feature-table summarize \
  --i-table amrfinderplus_abundances_ft.qza \
  --o-visualization amrfinderplus_abundances_summary.qzv

Your visualization should look similar to this one.