Figure 1. This diagram outlines relationships amongst different metagenomic CRISPR detection methods.

CRISPR detection can be performed either using specified direct repeats (reference-guided detection) or without prior knowledge of direct repeat sequences (de novo detection). De novo detection searches raw metagenomic reads for direct repeat sequences of the appropriate length and spacing (i.e., 25–60 bp long repeats with 25–60 bp spacers between them). De novo detection techniques either detect spacers in reads only (MinCED) or assemble reads into arrays (Crass). Reference-guided CRISPR detection, on the other hand, searches reads for user-specified direct repeat sequences, and extracts spacers from between direct repeat sequences identified in reads containing direct repeats. While the query is user-specified, general strategies for generating a query include using direct repeats found in assembled metagenomic contigs with CRISPR array detection tools (e.g., PILER-CR) or direct repeats found in genomic CRISPR arrays (e.g., those found in microbial genomes included in CRISPRdb) that might be expected based on taxonomic profiles. An example of the latter strategy would be searching for known genomic Streptococcus pyogenes direct repeats if Streptococcus pyogenes is found in the metagenome’s taxonomic profile.