What it does
Start from a PDB ID, a file, a protein's name or just a sequence. Ask for a single step or the whole chain: each step is checked and priced before it runs, and each result feeds the next.
Predict
A structure from its sequence, alone or folded together with a small molecule, DNA or RNA, with an estimate of how tightly the molecule binds. Two predictors: one for complexes, one for designed proteins.
Find the pocket
Pockets are found from the protein's shape, free. For a second opinion, a deeper search maps where probe atoms bind. Or simply name the site, or describe it in words.
Dock
Small molecules in every likely charge and stereo form, peptides from their sequence, and partner proteins, zinc sites included. What holds any pose is one click away.
Screen
Up to a million compounds, from your own library, a link or one of ours. The best hits are docked again more carefully, ready to simulate.
Simulate
Molecular dynamics in independent replicas, protonated for your pH, with every molecule parameterized. Checkpoints mean an interruption resumes instead of starting over.
QM/MM
A quantum region exactly where the chemistry happens: a ligand on its metal, a proton changing hands, a bond forming or breaking.
What comes back
Every simulation comes back with a plain answer, charts, a playable trajectory, an interaction map, a PDF report with its methods, and the files to rerun it yourself; every other step gets a report of its own. The contacts of any structure or pose are free, and everything in the app is open to your scripts with an API key.
About the demo
The demo on the front page asks where JQ1 binds a protein known only by its sequence, and whether it stays bound over 100 ns. ConformerLab lays out the steps, predicts the structure (9.6 credits), finds the pockets, docks JQ1 into the best one (6 credits), simulates the complex in three replicas (600 credits) and plays the result: JQ1 stays in its pocket, held by Asn140.
The steps, wording and prices are exactly what ConformerLab shows for this question. The five models and their confidences are from a real prediction of this sequence, whose best model lies 0.46 Å from the crystal structure (PDB 3MXF). Made for this page: the pockets, JQ1's pose and the footage use that crystal structure instead of the predicted model; ConformerLab's own pocket search and docking put JQ1 0.51 Å from where the crystal holds it, and the footage is 0.6 ns of that pose in implicit water, not the 100 ns run.
See the same study run for real, from the sequence, with its report and files.
Tested against crystal structures
Each step was tried, the way your studies run, on a case whose answer is already known. These are single checks from our own testing in October 2026, not yet a published benchmark.
Predicting: 0.34 Å
Ubiquitin predicted from its sequence alone: the best of five models lies 0.34 Å (Cα, residues 1–70) from the crystal structure, PDB 1UBQ. BRD4's first bromodomain: 0.46 Å from PDB 3MXF.
Docking: 0.51 Å
JQ1 docked back into BRD4's first bromodomain (PDB 3MXF, JQ1 taken out): the best pose lands 0.51 Å from where the crystal holds it.
Finding the pocket: first
Carbonic anhydrase II (PDB 3HS4): the deeper pocket search ranks its zinc active site first of all the pockets it finds.
Folding together: 13 nM
BRD4 and JQ1 folded together from a sequence and a SMILES: JQ1 lands in the acetyl-lysine pocket, with a binding estimate of 13 nM where about 50 nM is measured.
How a study goes
1. Ask
The way you would ask a colleague. Mention a PDB ID, a protein's name or a sequence, or upload a file; the molecules and peptides you mention are picked up from your words.
2. Agree the plan
The AI planner proposes the steps and settles them with you, one decision at a time, while a checker tests every setting against published practice. Every step carries its price; change anything and the price updates.
3. Run it and read it
Approve, and it runs. Long runs checkpoint and pick up where they stopped. You get the answer, the evidence behind it and the files to rerun it yourself.
Your data
Your studies, files and results answer only to your account; to anyone else they do not exist. A structure file you open is read in your browser, and uploaded only when you approve a run, which reaches only its own files.
Your questions go to the AI model behind the planner, and a lookup you ask for goes to the public database that answers it. There are no analytics and no advertising.
A run's files are kept free for 30 days, longer if you choose. Download everything, or delete your account, at any time. The privacy policy names every kind of party that receives anything, and what it receives.
Pricing
Everything is paid in credits, bought ahead: $0.025 each as you go, as little as $0.02 in a pack. Every quote is free. So is planning by rules, and the AI planner's first 40 credits each month. You approve the exact price before anything runs, and a run never costs more than its quote: you pay for the compute your results actually used.
Runs
Prediction, docking, screening, simulation and QM/MM: one credit a minute of compute, so an hour is 60 credits. Each is quoted before it starts and never costs more.
The AI planner and assistants
The first 40 credits each month are free. After that each reply is charged from your credits, and shows what it cost. The AI planner plans with you once your email is confirmed; planning by rules is always free.
Reports
Free up to 10 pages. A longer report you build costs 2 credits a page past those, once.
Storage
A run's files are kept free for 30 days. Keeping them longer is 1.8 credits per GB a month (about $0.045).
Packs
$20 for 900 credits ($0.022 a credit, 11% off); $100 for 4,700 credits ($0.021 a credit, 14% off); $200 for 10,000 credits ($0.02 a credit, 20% off). Or pay as you go at $0.025 a credit, 40 credits a dollar, any whole amount from $20 to $5,000.
What studies cost
A typical simulation (a 200-residue protein, 100 ns, three replicas) is quoted at about 700–1,300 credits: $17.50–32.50 paying as you go, or $14–26 with the $200 pack.
Shorter steps cost little: a small protein's structure (five models) about 9 credits ($0.23); docking a molecule 6 ($0.15), a peptide 12 ($0.30), a protein 30 ($0.75); the deeper pocket search 3 ($0.08); exact contacts 5 ($0.12); a fast screen of a million compounds about 2,200 ($55). Dollars at the pay-as-you-go price.
Refunds and payment
A run that fails on our side is refunded in full, automatically, and unused paid credits can be refunded within 30 days (see the refund policy). Prices are in US dollars, with sales tax or VAT added at checkout where it applies. Credits are sold by our reseller, Lemon Squeezy, the merchant of record, which takes the payment and sends your receipt. Need a purchase order or an invoice? Write to info@conformerlab.com.