Features

Qiviut docs

Qiviut predicts the 3D structure of proteins, DNA, RNA and the ligands they bind. Underneath is MuskOx, the computational kernel Auxfield built: physics, not a trained model.

Every kind of entity

Protein, DNA and RNA chains, ligands as SMILES or a CCD code, and twelve common ions. Up to 20 chains per job.

A sequence ruler

Sequences show in blocks of ten with positions, and any letter that isn't valid is marked where it sits. Paste a FASTA file with several records and each becomes its own card.

Quoted before it runs

The price shows as you build a job. Credits are held when it starts and charged only when the structure is delivered.

Structure view

Turn the model in the browser, as a trace or as lettered residues, colored by position or by chain, with ligand poses in place.

Binding sites

The residues each ligand touches, marked on the sequence and in the structure.

Deterministic

The same input and databases give the same structure, every run.

Yours to use

Commercial use is included. Your jobs are private to your account.

An API

Submit and fetch jobs from your own code.

Guide

Run a prediction

1. Add your entities. Each card is one entity. Pick its type from the menu at the start of the card; switching type keeps what you typed for each kind. Paste a sequence, a SMILES string or a CCD code, or choose an ion.
2. Set copies. Copies repeat a card: a homodimer is one protein card with 2 copies. Every copy is a chain or a ligand pose of its own, and each is counted in the quote.
3. Check the ruler. Click away from a sequence to see it as a ruler. Red letters aren't valid for that type: proteins take the twenty standard amino acids, DNA takes A, C, G and T, and RNA takes A, C, G and U.
4. Organize. The ⋮ menu on each card duplicates, moves or removes it, and the arrow collapses it.
5. Predict. The quote is at the bottom. Press Predict structure; the job opens and updates on its own until it's done.

Results

Read a result

Style

Trace draws the backbone with depth: near strands thicker and stronger. Residues draws each residue as a lettered bead, best for short chains.

Color

Position shades from the start of each chain to its end. Chain gives every chain its own color.

Binding sites

Residues with any atom within 4.5 Å of a ligand atom in the model. Hover a ligand to pick out its site.

Downloads

The model as PDB, and one SDF per ligand pose.

Clone and reuse

Copies a job's inputs into a new prediction, to change and run again.

API

One call per structure

Send the chains with your key, poll the job, and read back the structure.

# Submit a job
curl https://qiviut.auxfield.com/v1/jobs \
  -H "Authorization: Bearer qv_…" \
  -H "Content-Type: application/json" \
  -d '{"title": "demo", "sequence": ["GYDPETGTWG"], "ligand": ["ATP"]}'

# Poll until "status" is "done", then read "pdb"
curl https://qiviut.auxfield.com/v1/jobs/job_… \
  -H "Authorization: Bearer qv_…"

Chains

sequence, dna and rna each take a list of sequences. Or send chains, an ordered list of {"type": "protein", "sequence": "…"}.

Ligands

ligand is a list of SMILES strings or CCD codes, one entry per copy.

Response

Submitting returns the job id and its quote. A finished job carries pdb, up to five ranked models, and ligands, one SDF per pose.

Keys

Keys begin qv_. Key management on this page is on its way; until then, ask us for one at hello@auxfield.com.

Pricing

Pay for what you fold

A job costs credit per residue across all chains, plus per ligand or ion copy, with a minimum of . Credits come in packs, with a volume discount on larger packs. New accounts start with free credits. Each month, an account below credits gets a free top-up back to .

One-time payments through Paddle, no subscription. Credits don't expire. Prices in US dollars; tax is added at checkout where it applies. Every job is quoted in credits before it runs.

FAQ

Questions