Design macrocyclic peptide binders against any target structure
Select an interface on a disease-relevant protein target and a known interacting fragment. Get back ranked, disulfide-constrained peptide binders, cyclizable on demand, hosted, GPU-powered, with zero infrastructure to stand up.
The pipeline combines leading open-access AI models with proprietary tools. You provide the target's PDB structure, the interacting fragment (PDB or sequence), and the interface hotspots; the system builds peptide sequences matched to the interface topology. Candidates are ranked by a composite score combining standard AI structure-prediction metrics (pLDDT, ipTM) with a physics-based ΔG estimate. Benchmark a known binder alongside your candidates to calibrate what a strong composite score looks like for your target. A team of AI co-scientists is on hand to help refine designs and interpret the scores.
AI Generated predictive output based on existing training data used in the industry.
Anchored to a real, established interaction, not a fixed disease area. Bring any target structure and a fragment already known to engage it; the method itself doesn't care what the target is.
Design against the target's surface, the fragment, or both in one run, independently validated and ranked, and you only pay compute for the side(s) you select.
Every candidate clears developability screening, independent structural validation, and affinity scoring before it reaches your shortlist, and the ranking says so when the numbers aren't strong enough to trust.
Every run is backed by a panel of specialist AI co-scientists in structural biology, chemistry, and validation, who argue through the results the way a real cross-functional lab would, not a single black-box model grading its own homework.
Docking geometry, backbone generation, sequence design, co-folded validation, and scoring; one orchestrated run per job, not a chain of manual steps.
Choose A100 or H200. Jobs stream live progress and typically finish in minutes to a few hours; no cluster to provision, no queue to manage.
Integrated directly into your annual license. Preview estimated compute resources before running complex generation jobs, backed by itemized usage logs for full internal cost control and departmental accounting.
Uploaded structures and results are encrypted and scoped to your account under a zero-data-retention policy, never used to train any model.
From a target to a defended shortlist.
- STEP 01
Anchor to a known interaction
Provide a PDB code or structure file for your target, and a sequence for a fragment already known to bind it. Select hotspot residues directly on the interface.
- STEP 02
Set the design constraints
Configure the macrocycle: ring length, any fixed positions, and how many candidate backbones to sample per side.
- STEP 03
Get a validated, ranked shortlist
Watch the run progress live. Each candidate arrives with a structure, a sequence, structural and interface confidence, and a predicted binding free energy, flagged honestly wherever the numbers don't hold up.
- STEP 04
Discuss, refine, and challenge the results
Bring your shortlist to the AI co-scientist team. Push back on a score, ask why a candidate ranked where it did, or send favorites for relaxation or bicyclization; refined and re-evaluated on demand, never applied automatically.