Data corrections

The binomial parser treated any third word starting with a capital as a nomenclatural authority. NCBI's placeholder taxon uncultured bacterium PAD7 (taxid 48476) therefore registered the index key uncultured bacterium, and every SILVA record whose organism name began “Uncultured bacterium …” was bound to that single taxon.

1,233,567 SILVA records — 55.4% of the entire taxmap — were funnelled into four placeholder taxa. Taxid 48476 alone absorbed 980,251 of them. For scale, Escherichia coli has 16,456. The same defect produced bindings for Bacillus sp., Pseudomonas sp. and similar non-species.

Removing those bindings drops SILVA from 29,830 taxa to 18,669. Nothing real was lost: the removed cells pointed at NCBI names that are themselves placeholders. A SILVA column that is smaller and correct is worth more than one that is larger and wrong.

NCBI writes provisional names as Candidatus Nanopelagicus abundans, where the second word is the genus. The parser read it as genus + species and then failed, so 8,869 species-rank Candidatus names produced 5 index keys — and those 5 were wrong, keyed on “candidatus <genus>”, which let them attract records belonging to other organisms.

Candidatus names are a large and growing share of environmental microbiology. Both the index and the search box now strip the prefix, so Candidatus Pelagibacter ubique and Pelagibacter ubique reach the same taxon. Fixing only the index — as this page briefly claimed — would have left every one of them unfindable by its own scientific name.

GTDB records a genome against whichever NCBI taxid fits best, and that is often a strain or subspecies rather than a species — 25,520 distinct strain taxids in R220 alone. Those rows were silently discarded.

They are now resolved to their species, preferring GTDB's own ncbi_species_taxid column, then a walk up the NCBI tree, then a merged-taxid redirect. All three GTDB releases now bind 100% of their genome rows (R220: 596,859 of 596,859, up from 90.7%).

Cells used to join their three most common names regardless of support. NCBI taxid 485 (Neisseria gonorrhoeae) has 43 Greengenes2 genomes called N. gonorrhoeae and one called Bifidobacterium vaginale — and rendered as “Neisseria gonorrhoeae / Bifidobacterium vaginale”, which reads as a genuine split rather than one misannotated record.

Only names meeting the support threshold now appear in the cell. The rest are still recorded and shown under the taxon's full name list — nothing is hidden, it simply is no longer asserted. Every cell also carries its genome count, so “current” can be told apart from “too few genomes to check”, which is the case for most taxa.

This is the most visible consequence, so it is worth being exact about. A taxon gets a page only while at least one database still holds a record for it. 11,235 of the 11,285 that lost theirs had SILVA as their only source — and that SILVA cell was one of the placeholder bindings described above. Their page was built on a record that was never about them.

The remaining 50 are a different case: genomes are now attributed to the species Greengenes2 and GTDB themselves assign them to, rather than to whatever NCBI taxid the genome record happened to carry, and for those taxa every genome moved elsewhere. Examples include Couchioplanes azureus, Xanthobacter versatilis, Planomonospora antibiotica and Alcaligenes phenolicus.

Nothing here is a judgement about whether a species is valid. These are validly published names; the four databases this site reads simply hold no current record against them. Searching for one now says so, instead of showing a page assembled from a record belonging to something else.

GTDB R214 and SILVA 138.1 are now selectable. SILVA 138.1 is interesting for a specific reason: it differs from 138.2 in zero organism names and in 15,697 lineages — it is where Proteobacteria becomes Pseudomonadota and Firmicutes becomes Bacillota. It was previously excluded because, while only names were compared, the two releases were indistinguishable.

Greengenes2 2022.10 joins through GTDB R207. It is built on the R207 backbone, and this site holds R207 for that reason. Joining it through a different GTDB release would produce answers that look fine and are wrong, so 2022.10 was absent until R207 landed (29 October 2026).

If you cited a result from this site before 27 July 2026, re-check it. The most likely difference is a SILVA cell that used to show a name and now shows none — in almost every such case the old name was a placeholder binding, not a record about your organism.

Current releases: NCBI Taxonomy taxdump · 2026-10-23 · GTDB R232 · SILVA 138.2 · Greengenes2 2024.09. Full method: how matching works.