KSGP 3.1: improved taxonomic annotation of Archaea communities using LotuS2, the genome taxonomy database and RNAseq data

Taxonomic annotation is a substantial challenge for Archaea metabarcoding. A limited number of reference sequences are available; a substantial fraction of phylogenetic diversity is not fully characterized; widely used databases do not reflect current archaeal taxonomy and contain mislabelled sequences. We address these gaps with a systematic and tractable approach based around the Genome Taxonomy Database (GTDB) combined with the eukaryote PR2 and MIDORI mitochondrial databases. After removing incongruent, chimeric and duplicate SSU sequences, this combination (<i>GTDB+</i>) provides a small improvement in annotation of a set of estuarine Archaea Operational Taxonomic Units (OTUs) compared to SILVA. We add to this a collection of near full length rRNA sequences and the prokaryote SSU sequences in SILVA, creating a new reference database, KSGP ( <b><i>K</i></b> arst, <b><i>S</i></b> ilva, <b><i>G</i></b> TDB, and <b>P</b>R2). The additional sequences are (re-)annotated using three different approaches. The most conservative, using lowest common ancestor, gives a further small improvement. Annotation using SINTAX increases Class and Order assignments by 2.7 and 4.2 times over SILVA, although this may include some "lumping" of un-named and named clades. Still further improvement can be made using similarity based clustering to group database sequences into putative taxa at all taxonomic levels, assigning 60% and 41% of Archaea OTUs to putative family and genus level taxa respectively. GTDB without cleaning and GreenGenes2 both perform poorly and cannot be recommended for use with Archaea. We make the GTDB+ and KSGP databases available at ksgp.earlham.ac.uk; integrate them into a metabarcoding pipeline, LotuS2 and outline their use to annotate Archaea OTUs and metatranscriptomic data.