PriSeT: Efficient De Novo Primer Discovery

Marie Hoffmann; Michael T. Monaghan; Knut Reinert

In metagenomics, the DNA of hundreds to thousands of species from a single environmental sample is processed in parallel and one goal is to identify all species from a sample. Samples contain many species in what can be very heterogeneous communities. Their identification is usually done via DNA metabarcoding which combines polymerase chain reaction (PCR), nextgeneration sequencing (NGS) and identification via DNA barcodes, i.e. short amplicons that are matched against a reference database. In contrast, single DNA probing and assembly would be more precise, but are for many reasons not feasible, e.g., the majority of microorganisms cannot be clonally cultured -a necessity for assembling reference genomes (Rappé & Giovannoni, 2003) . Although the decreasing costs of NGS makes DNA metabarcoding progressively feasible, identification of most microeukaryotes, such as plankton, continues to be done via morphology. This still outperforms molecular methods in some aspects, like abundance estimation, differentiation of life stages, or observation of teratological forms. But metabarcoding of environmental DNA (eDNA) is a far more sensitive method of species detection compared to tradi-tional methods (Smart et al., 2015) provided that a suitable marker is chosen. In the search for a marker, we know that evolutionary pressure is not consistent within genomes. There are regions that vary even within species, or on the other extreme, are conserved for whole clades. For an effective delineation of taxa using DNA metabarcodes, we are interested in regions that are similar or identical to individuals of the same species, but distinctive from other species. Such regions are ideal barcodes, but they need to be surrounded by conserved regions that contain suitable binding sites for primers. Primers are short oligomers, ideally 16 to 25 bases long, which determine the copy starting point of the template. We need flanking conservation, because we batch-process all extracted DNA at once and can add only one primer pair per PCR run 2 . The most frequently applied type of NGS in DNA metabarcoding is the paired-end approach as it allows for longer sequences and error correction for the low quality ends of reads. The two types of primers operate independently, one binds on the forward and the other on the reverse strand of the template. Both are elongated enzymatically to produce a complementary copy. In each PCR cycle the number of amplicons is nearly duplicated. After about 25-35 cycles the reaction is stopped and the amplified DNA amplicons provide a signal for the sequencer machine that is separable from the noise of the original DNA material. After digitization of the reads, an analysis pipeline combines trimming, quality filtering, denoising, and read merging with the aforementioned error correction. Identical or similar sequences are then clustered into operational taxonomic units (OTUs). The underlying hypothesis is that each OTU corresponds to a taxon in the tree of life. An alternative approach is that each sequence variant is analysed rather than being clustered into OTUs. The art herein is to discriminate sequencing errors from variants. A noteworthy denoising algorithm is the Divisive Amplicon Denoising Algorithm 2 (DADA 2) (Callahan et al., 2016) . As described earlier, not all barcodes are specific to a single species and our metabarcoding pipeline will most likely produce OTUs uniting barcodes from taxonomically distinct lineages, even with simil