Taxonomic classification

Let’s now find out what the taxonomic composition of our samples is. There are at least two ways to do this. We can classify the reads directly to get an early glimpse into the taxa within, or we can classify the MAGs we recovered to get a more accurate picture of the community members. Let’s try both approaches. In this section we will use Kraken 2 as the classifier of choice.

Database preparation¶

Before we perform the classification, we need to get the reference database. For this tutorial, we will use the “standard” Kraken 2 database with the smallest memory footprint so that it’s easier to run on a laptop.

mosh annotate build-kraken-db \
    --p-collection standard8 \
    --o-kraken2-db kraken2-db.qza \
    --o-bracken-db bracken-db.qza \
    --verbose

Read-based classification¶

Let’s start with read-based classification. We can use the classify-kraken2 action to perform the classification and then correct the read counts to account for the genome size and database composition using the estimate-bracken action.

mosh annotate classify-kraken2 \
    --i-seqs reads.qza \
    --i-db kraken2-db.qza \
    --p-threads 4 \
    --p-memory-mapping \
    --o-reports kraken2-reports-reads.qza \
    --o-outputs kraken2-hits-reads.qza \
    --parallel-config parallel.config.toml \
    --verbose

mosh annotate classify-kraken2 \
    --i-seqs reads.qza \
    --i-db kraken2-db.qza \
    --p-threads 4 \
    --p-memory-mapping \
    --o-reports kraken2-reports-reads.qza \
    --o-outputs kraken2-hits-reads.qza \
    --verbose

Before we can visualize the abundance of the identified taxa, we need to correct the read counts slightly to take into account their genome sizes and database composition. For this purpose, we can use the Bracken tool which uses statistical methods to re-estimate read counts to account for those factors. Run the action below to perform that correction:

mosh annotate estimate-bracken \
    --i-kraken2-reports kraken2-reports-reads.qza \
    --i-db bracken-db.qza \
    --p-read-len 150 \
    --o-reports bracken-reports.qza \
    --o-taxonomy bracken-taxonomy.qza \
    --o-table bracken-table.qza \
    --verbose

Now, we can use the familiar barplot action to plot the taxa abundance in our samples:

mosh taxa barplot \
    --i-table bracken-table.qza \
    --i-taxonomy bracken-taxonomy.qza \
    --o-visualization bracken-barplot.qzv \
    --verbose

Your visualization should look similar to this one.

MAG-based classification¶

It is also possible to classify the MAGs we recovered using Kraken 2. We can use the same classify-kraken2 action as before but pass the dereplicated MAGs we got in the previous section as input.

mosh annotate classify-kraken2 \
    --i-seqs mags-derep.qza \
    --i-db kraken2-db.qza \
    --p-threads 4 \
    --p-memory-mapping \
    --o-reports kraken2-reports-mags.qza \
    --o-outputs kraken2-hits-mags.qza \
    --parallel-config parallel.config.toml \
    --verbose

mosh annotate classify-kraken2 \
    --i-seqs mags-derep.qza \
    --i-db kraken2-db.qza \
    --p-threads 4 \
    --p-memory-mapping \
    --o-reports kraken2-reports-mags.qza \
    --o-outputs kraken2-hits-mags.qza \
    --verbose

We got two new artifacts: FeatureData[Kraken2Report % Properties('mags')] and FeatureData[Kraken2Output % Properties('mags')]. The first one contains the Kraken 2 report: a tree-like representation of all the identified taxa. The second one is a list of all MAGs with their corresponding identified taxa. To convert those into a more QIIME-like taxonomy, run the following action:

mosh annotate kraken2-to-mag-features \
    --i-reports kraken2-reports-mags.qza \
    --i-outputs kraken2-hits-mags.qza \
    --o-taxonomy mags-taxonomy.qza \
    --verbose

A natural next step would now be to estimate the relative frequencies of those taxa in our samples. It is, however, a little bit more complicated in the case of MAGs, since we would need to take into account how many reads map to each of our MAGs and from there try to infer their abundance. Head to the next section to learn how to do that.