Genome Taxonomy Database

Contents of the database

Each release contains:

• Taxonomy tables containing the assignment of all included genome assemblies to the phylum-to-species taxonomy. (One per domain.)
• Files containing the metadata given to each genome assembly, including original taxonomy from NCBI, original strain identifier, GTDB taxonomy, quality estimates, and presence of important genes (tRNA and rRNA). (One per domain.)
• Species tree Newick files containing the species-representative genomes (1 per species), built as described in the previous section. (One per domain.)
• For species-representative genomes:
  • alignments of marker genes identified from these genomes
  • file containing one 16S rRNA sequence from each species
  • tarballs containing amino acid and nucleotide versions of all predicted genes in these genomes
  • tarball containing the full contents of all these genomes
• For all genomes that pass quality check:
  • alignments of marker genes identified from these genomes
  • file containing all 16S rRNA sequences identified from these genomes
• Auxiliary files; see the full FILE_DESCRIPTIONS.txt。

The web interface displays a tree based on the taxonomy (not the entire Newick file), down to the genome assembly level. Each genome assembly has a page detailing its metadata and a history of how it's classified in each GTDB release. There is a search functionality.

Effects on the accepted taxonomy

GTDB "has now become an important resource for prokaryotic taxonomy". Both its species tree and elements of its methodology are used by taxonomists to improve the current, accepted taxonomy under the Prokaryotic Code. For example, a taxonomist may make references to the GTDB tree on top of their own phylogenetic tree to further support a taxonomic proposal.

There has been even more ambitious proposals to import large parts of the database into the accepted taxonomy. A 2022 article in the IJSEM, written by third-party authors, proposes to assign names based on meaningless Latin syllable to over 65 thousand GTDB taxa, though none of these names have made their way into the LPSN. A 2023 article by the GTDB team proposes to import 223 higher-order taxa into the Prokaryotic Code system and 49 under the SeqCode system. Many of the names published under the Prokaryotic Code have already been validated. (The SeqCode requires registration of the names for valid publication, which has also been done.)

Cyanobacteria

Unlike other bacteria, cyanobacteria are governed by a separate nomenclature code, the Botanical Code (ICNafp), and has a rich collection of taxa described morphologically, before the advent of molecular methods. Considerable work has been done at the higher taxonomical ranks (orders and families) to harmonize them with the results of molecular phylogeny, but problematic taxa remain at the lower levels. One representative example is the genus Synechococcus (and the broader Synechococcales), where distantly-related taxa remain lumped together. A 2020 article proposed using GTDB's methods to split this taxon into five orders, but it did not make an effect on the accepted taxonomy as the names were not validly published.

The Botanical Code treats names validly published under the Prokaryotic Code as valid, but the converse was not true until 2022, when the Prokaryotic Code reciprocated. Using some cyanobacterial genera names published under the Botanical Code, the aforementioned 2023 GTDB article proposed 10 cyanobacterial higher taxa to be accepted under the Prokrayotic Code.

The cyanobacterial community revised its higher-level taxa in 2023 independently of the GTDB, making use of a combination of phylogenomic and 16S trees. Major differences compared to GTDB include:

• GTDB takes into account genera and species never validly published and builds higher-order taxa such as "Leptococcales" from them. The corresponding clade does not have a name in the 2023 taxonomy.
• Many cyanobacterial taxa are justified by morphological differences that do not necessarily correspond to a large evolutional divergence. GTDB exhibits lumping of such taxa due to RED normalization. For example, "Cyanobacteriales" corresponds to many orders in the 2023 taxonomy: Desertifilales, Geitlerinematales, Oscillatoriales, Spirulinales, Coleofasciculales, Chroococales, Gomontiellales, Chroococcidiopsidales, Nostocales.

More broadly, the nomenclatural situation of cyanobacteria remains disordered enough to cause several confusing assignments in GTDB requiring manual adjustment. For example:

• Some cyanobacterial taxa are known to be nested in another and still provisionally kept for having "unique and distinct" features. GTDB does not respect this distinction, leading to reassignments. Results can be confusing when the genera to be merged have species that share the same specific epithets. An example is described at Aphanizomenon flos-aquae § Taxonomy and diversity.
• Some genera do not have a genome for its type species, nor a species closely allied to the type. This can cause a surprisingly different concept of a genus compared to the accepted taxonomy as GTDB only uses the genome, never other sources such as 16S rDNA. For example, the GTDB concept of "Limnothrix" is defined around a Limnothrix rosea, which is known to more properly belong in Picosynechococcus.

(Both of the above may happen for any group of prokaryotes, but cyanobacteria are particularly affected due to a higher scarcity of genomes and a historical delay in adopting molecular-based methods.)