CRUX: Creating Reference libraries Using eXisting tools

Taxonomic assignments of metabarcoding reads requires a comprehensive reference library. There are preexisting high quality reference libraries that are compatible with 16S and 18S metabarcode primers (Glöckner et al., 2017, DeSantis et al., 2006). There are few other comprehensive metabarcode specific reference libraries. Using CRUX: Constructing Reference libraries Using eXisting tools (CRUX) we are able to construct custom reference libraries given a primer set and several publicly available databases.

CRUX generates custon reference libraries by first running in silico PCR (ecoPCR; Ficetola et al., 2010) against the EMBL standard nucleotide sequence database (Stoesser et al., 2002) to generate a seed library of reads with unique taxon identifiers. Other nucleotide databases can also be used in this step (e.g. genbank, custom fasta files, etc.). CRUX verifies that the seed library reads match the amplicon by checking for the correct primer regions and trimming them using cutadapt (Martin, 2011).

Because many sequencing records are deposited to Genbank (Benson et al., 2012) with the primer regions removed from the read, we BLAST (Camacho et al., 2009) the seed library against the NCBI nucleotide blast database (ftp://ftp.ncbi.nlm.nih.gov/blast/). CRUX runs blastn twice. The first blastn run only accepts full length reads (e.g. the same length as the Reference) and then de-replicates the resulting fasta files by NCBI accession version number. The second blastn run accepts reads up to 70% of full length (because many hits to metabarcodes do not cover the entire read length). The resulting reads are sorted by length and de-replicated so that only the longest version of a read is retained.

A corresponding taxonomy identification file (superkingdom, phylum, class, order, family, genus, species) is generated from the fasta formatted blast output using entrez-qiime (https://github.com/bakerccm/entrez_qiime), the NCBI’s taxonomy dump and map of association between taxonomy and accession version numbers (ftp://ftp.ncbi.nlm.nih.gov/pub/taxonomy/). Because NCBI taxonomy is not always complete (e.g. uncultured organisms, organisms with unknown taxonomy, etc.) CRUX generates two sets of reference files: filtered and unfiltered. The filtered reference files exclude reads with the following in their taxonomic path: 'unassigned', 'unknown', 'uncultured', 'environmental', 'sample', or 'NA;NA;NA;NA'.

The downstream metabarcoding Anacapa pipeline (https://github.com/limey-bean/Anacapa) runs Bowtie2 (Langmead et al., 2009) to assigns reads to these references databases and generate a corresponding taxonomy file. Thus the final step is to generate Bowtie2 formatted data index libraries.