Getting started
Start from the protein you already have, run one small job on it, and read what comes back.
-
Start from what you already have
Two things decide which tool you need: what is on your bench right now, and what you want back. A 3D structure of a target you want something to stick to. A structure with no sequence yet, and you need sequences that fold into it. A sequence, and you want to see its predicted shape or whether it will survive being made. Each of those is a different tool, and you do not need to know any model names to find it — the home page groups the tools by that question, and the full tool list puts them side by side with typical runtimes.
-
Open the tool page and look at the form
You do not need an account to look. The tool page shows the exact inputs the run asks for, and a cost estimate that updates live as you change the settings. Most binder design tools want three things: your target structure as a PDB or mmCIF file, which chain in that file to aim at, and a few hotspot residues — the specific amino acids on the target's surface you want the new binder to touch. If you do not know yours yet, score your target's surface with Epitope Scout first — it is free, and it is the step before any of this, not a fallback for when a paid run comes back empty. Structure prediction tools want a sequence instead.
-
Start the tool's named pilot
Most tool pages carry a pilot card: a named first run on your own target, sized to tell you whether the setup works before you spend on a large one. It states what the run is for, what you need in hand, and roughly what it costs before you commit anything. Load these settings fills the form with exactly those values — you supply your own structure and hotspots on top. A few tools have no pilot card, because their smallest possible run is already the only run they do; on those, just submit the form.
-
Sign in and submit
Submitting is the first point that needs an account. Billing is a USD wallet and that balance is your spend cap — no subscription, no monthly minimum, and a new account starts with $15 already in it, which covers every pilot on the site. Before you confirm you see the estimated cost and what your balance will be afterwards; a hold is placed against the estimate while the job runs and the final charge settles against the compute actually used. See pricing for how holds and top-ups work.
-
Read the result
The job page streams progress while the GPU works, then hands back your candidates ranked by score. Open any one of them in the 3D viewer, and export the score table as CSV or the sequences as FASTA. Each score has a rough cut-off above which a design is worth taking further — the guide for the tool you ran spells out which number to sort on and where its cut-off sits.
-
Scale up, or switch tools
If the pilot produced anything that scored well, clone the run and raise the design count — the setup is proven, so the only thing missing is volume. If nothing scored well, the setup is the usual culprit rather than the model: the wrong chain, the wrong patch of surface, or a target that is genuinely hard to bind. Score the target's surface first with Epitope Scout, or move to a tool built for the kind of binder you need.
-
Take the shortlist to the bench
Star the designs worth pursuing to build a shortlist, then hand it off to a Binder Pilot or AI Binder Sprint for wet-lab validation.
What to expect from a pilot: about 8 designs gives a real first signal, because roughly 1 in 5 designs passes the in-silico filter — the computational quality check the tool applies before anything reaches you — on a tractable target. Start small, read the scores, then scale the run up once you trust the setup.