We gathered every public 18S, 28S and COI sequence for named tardigrade species, screened out records that don't match their own species, and built one tree of 412 species. The outer ring is the official classification. Where DNA and names agree, each family's colour block is solid; where they don't, it breaks apart.
Scroll or pinch to zoom · drag to move · hover a species · click a disagreement on the right to light up its species
ArthrotardigradaEchiniscoideaApochelaParachela · the outer ring shades each family in its order's colour · a filled dot on a branch means strong support (SH-aLRT ≥ 80 and ultrafast bootstrap ≥ 95)
We downloaded every tardigrade 18S rRNA, 28S rRNA and COI record in GenBank and BOLD, matched each name to the world checklist, and kept one sequence per marker per species. A screen compared every record with the other records of its species from other studies; records that disagree with their own species' records from other studies, or match a different genus far better, were set aside, 176 in all, each with its reason. Most are hidden lineages under the names of species complexes such as Milnesium tardigradum and Macrobiotus hufelandi; one is a 2026 batch in which the same Macrobiotus sequence was filed under three unrelated species. We then built a tree from each gene alone, to see which disagreements more than one gene supports. The markers were aligned separately (MAFFT), joined, and a maximum-likelihood tree inferred with IQ-TREE, with the root between the two classes, as outgroup tests confirm.
About a quarter of described species have any of these sequences; six families and 60 genera have none. So the tree is a map of what has been sequenced as much as of how tardigrades are related, and some disagreements come from a single marker or a single study. Targeted studies with more genes (for example Stec et al. 2021 on Macrobiotidae; Gąsiorek & Michalczyk 2020 on echiniscids; Guil et al. 2019 across the phylum) are the reference for any one group. Species names follow Degma & Guidetti's world checklist (CC BY 4.0); sequences are public records in GenBank and BOLD.
build_moltree.py cuts each marker from its record's own annotation and keeps one sequence per marker per species. MAFFT (L-INS-i) aligns each marker, and IQ-TREE 3 infers the partitioned tree with 1,000 SH-aLRT and 1,000 ultrafast bootstrap replicates, plus one tree per marker. dna.js draws it here with d3 on a canvas.