From target to synthesis-ready candidates. We design, score, screen, and select — our wet-lab partners build and validate.
In short: we are a complete computational drug discovery platform. Bring us a target and we generate a candidate library, predict structure, screen for safety and patent problems, and rank what's left into a short, defensible build list. Wet-lab partners synthesize and validate; their results feed back into the engine.
This is a schematic of one step in the discovery pipeline — candidates in, screening, a ranked build list out. It illustrates the process; it is not prediction output for specific molecules.
Your team can keep designing with whatever generator it already uses. We don't ask you to switch. What we add is the layer after design: one consistent scoring and governance engine that every candidate passes through, regardless of where it came from — so the decision is made the same way every time, and the reasoning behind it is recorded, not just the outcome.
Eight stages, one auditable thread. The first six run inside the platform; the last two happen with your wet-lab partner, and what they learn comes back in.
You provide the biological target.
We generate a candidate library against it.
Boltz-2 — an open, third-party model we license and run, not one we built — predicts each candidate's 3D structure and its complex with the target, with interface confidence.
Our selection engine ranks every candidate that's left.
Biosafety (could it be misused as a biothreat) and freedom-to-operate (does it infringe existing patents) screening — fail-closed, meaning any candidate we can't clear is blocked, not waved through.
Ranked, auditable, sealed to a tamper-evident record.
Our synthesis and validation partners build it.
What comes back from the bench.
Closing the loop: wet-lab outcomes feed back into the design and scoring engine, sharpening every candidate library that follows — the platform gets better informed with every round, not just faster.
Schematic of the discovery pipeline. Stages 1–6 run inside the platform; stages 7–8 happen with a wet-lab partner and are not LillixBio outputs.
AI can now design more drug candidates than any company can afford to synthesize and test. Generating ideas is no longer the bottleneck — knowing which few are worth a wet lab's time is. LillixBio is the computational engine that makes that call, end to end, from a named target to a synthesis-ready build list.
Not a spreadsheet, not a score in an email — a sealed record. Every candidate on a build list carries its own provenance hash, biosafety verdict, and freedom-to-operate status. The list itself is sealed, with who sealed it and when, so a regulator, a board, or a funder can check it independently, without needing to trust us.
| Rank | Candidate | Source | Biosafety | FTO | Provenance hash |
|---|---|---|---|---|---|
| 1 | SAMPLE-0417 | RFdiffusion | Clear | Clear | cc9100a525a833c6a39ea0d7b8ebfb4d4d5e0f496f2e726041f0999bdd8c720c |
| 2 | SAMPLE-0298 | LillixBio design | Clear | Clear | f2c2bf821f4b1ad813b63703712ea6959882f479a61f978dc57b7b7a7e9a8485 |
| 3 | SAMPLE-0165 | ProteinMPNN | Clear | Clear | 502322879714c4dcbc6083043977d32b0e803ce6b8e03f360c9015aadaee5459 |
| 4 | SAMPLE-0512 | Customer pipeline | Clear | Clear | d687dc831f05daa7b276042f6f1c1e9e0e419e494c6384553ef52c3a56765898 |
| 5 | SAMPLE-0033 | RFdiffusion | Clear | Clear | 5fe1c037560a2dcff6c6e15a664d87045e440d78d11b5312d81f878c08a9dbec |
Recomputed live from the values shown above, using SHA‑256 in your browser — nothing is sent to a server. Candidate, source, biosafety, and FTO cells are editable: try changing one and watch the status change. Full scheme: how this is computed →
What this proves, and what it doesn't: a match means this record hasn't been altered since it was sealed. It does not prove who produced the list, and it can't stop someone who holds the full record from generating a different chain from scratch.
Illustrative example built to show the structure of the artifact. Candidate identifiers, hashes, source labels, and the sealer name are fabricated for demonstration and do not represent any real customer engagement, target, or molecule.
Same discovery engine, three ways to use it — sized to how each customer actually works.
The discovery engine at the center of the pipeline: designs or ingests candidates, predicts structure, screens for safety and patent problems, and ranks what's left into a short build list — with the reasoning for each pick attached.
See DECIDE ↓The clearance step every candidate passes before it reaches a build list, available standalone. Every candidate gets a clear yes-or-no, plus a report you can hand to a regulator.
See CLEAR ↓The operational layer of discovery: tracks what happens after a build list ships to the wet lab — tied to the same paper trail DECIDE already created.
See OPERATE ↓Each engagement is scoped around something specific you want built, and measured against whatever you're already using today.