← SkillSafe / Clade Desk

Your alignment and your tree, read properly before a reviewer reads them

Paste the FASTA alignment, the Newick string, and the few lines of notes that say what you ran. The page parses both files in your browser, draws the tree, and measures it — then the lane you pick turns those measurements into a review.

Both examples ship with a saved model run for every lane, so you can see the whole review — and the drawn tree — without signing in and without spending a credit.

nothing pasted yet

Drag a .fasta, .fa, .aln or .txt file here, or . Nothing is uploaded by dropping it — the file is read in this browser.

nothing pasted yet

Or drop a .nwk, .newick, .tre or .treefile here, or .

Optional, and the single highest-value thing you can add. The support threshold, the methods sentence and the figure caption all depend on it. Saying bootstrap: none is read as an answer, not as a gap.

How this works, and what it will not do

The free pass is a real reader, not a character count. The Newick tokenizer handles quoted labels that contain parentheses, bracketed comments, NHX annotations, negative and scientific-notation branch lengths, an internal support label sitting at the very end of the string, and IQ-TREE's combined SH-aLRT/UFBoot labels — which it reads as two series and grades against two different thresholds, because taking only the first number reports one metric under the other's bar.

The FASTA reader measures per-column occupancy, constant, variable and parsimony-informative sites, per-taxon gap and ambiguity fractions, characters outside the alphabet with their positions, duplicate identifiers, and sequences that are identical once gaps are removed. It then compares the tree's tip labels against the alignment's identifiers, and separates a genuinely missing taxon from one whose label differs only by an underscore — a rename and a redo are not the same problem.

Your notes are read for three states, not two: bootstrap: n/a, no bootstrap and bootstrap was not run all mean absent, while support: UFBoot, no gaps trimmed means present — the trailing "no" belongs to the gaps. Saying nothing at all is a third answer, and it is reported as unstated rather than guessed at.

Everything the browser measures is sent to the model as prescan_facts, and every one of those facts is reconciled against the model's answer afterwards. If the review skips a flag or invents one, this page says so under Reconciled against the free prescan rather than presenting a tidy answer. Severity moves with the mitigating facts already computed: weak support on a terminal sister pair is graded below weak support on the backbone, and a gappy alignment the notes say was trimmed with trimAl is graded below an untrimmed one.

What it will not do: run an aligner, infer a tree, or check your sequences against a database. It reads what you give it. Nothing you paste leaves this page except the lane you press, and the drawn tree, the measurements and the bundled examples are all free and work without an account.

Nothing to hand? Load the — an 18-taxon rbcL alignment with a ragged 5' end, a rogue taxon and a taxon missing from the tree — or the , an 8-taxon protein set that comes back clean. Both replay a saved run for free, in every lane.