Extra functions included in the workflow

Functions that are not directly related to CRISPR-Cas. That is, other bacterial genomics analyses that serve to better characterise and understand the input data.

Multilocus Sequence Typing

To facilitate comparison with other studies and estimate the genetic diversity within the input dataset, the workflow can assign Multilocus Sequence Types (MLST) using pyMLST, which can download databases from pubmlst.org.

Each genome is assigned its closest sequence type (ST) when possible.

Identification of antiviral defence systems

To identify different antiviral defence systems, we have included PADLOC. PADLOC uses a custom database to screen genomes for the presence of various systems. This information is written to a CSV file. We use this information to assess the prevalence of defence systems and calculate possible correlations with CRISPR-Cas.

Plasmid and virus prediction of input genomes

We have included two state-of-the-art tools for predicting the origin of contigs: chromosome, plasmid or virus (phage). These are geNomad and Jaeger. geNomad compares contigs to a marker database and classifies them using a neural network, aggregates these results and calculates probability scores for chromosome, plasmid and virus (these three sum up to 1). Jaeger uses deep-learning to classify sequences as viral or not. This can be used to identify 'free' phages as well as integrated prophages.

We map the identified CRISPR spacers back to the input genomes and use these plasmid/phage predictions to estimate the targets of the CRISPRs.