Design the molecule the disease needs.
OVIVO doesn't only judge molecules that already exist — it designs new ones. Discovery starts from the gaps a disease leaves in today's treatment landscape, generates candidate matter to fill them, and puts every candidate through the same unforgiving scrutiny as the rest of the platform — then validates the survivors through the full simulation chain.
When a route proves intractable, that's information — recorded and reused. We're as proud of the dead ends as the hits.
Most computational discovery starts at the binding site. We start at the disease.
The usual order is: pick a target, generate molecules that bind it, filter for drug-likeness, hand off to chemistry. Whether that target actually matters — in this disease, at this stage, given what's already being treated — comes later, if at all. The biology is an afterthought to the chemistry.
OVIVO inverts that. Discovery begins with the disease simulation: a map of which biological pathways the current treatment landscape covers, and which it leaves open. Targets are read from the gaps. Molecules are generated to fill them. And every candidate is judged not as a binding number in isolation, but as a drug effect in the context of the patient's biology — through the same chain that models everything else.
From a gap to a candidate — and back again
A directed chain, each stage feeding the next. A molecule that fails a gate doesn't proceed; one that passes is validated through the full simulation before it leaves the platform. Everything learned along the way flows back into the foundation.
-
1
Pathway gap analysis
Receives the disease simulation.
Maps which pathways the current treatment landscape covers and which it leaves open. Targets are identified from the uncovered mechanisms — not from a standing target list.
Sources (selection)Reactome · KEGG · Open Targets · DisGeNET
-
2
Target & structure
Receives a candidate target from the gap map.
Druggability assessment and structural characterization — experimental structure where it exists, predicted where it doesn't — then binding-site and pharmacophore modeling.
Sources (selection)PDB · AlphaFold · UniProt
-
3
Generative design
Receives the target and the modality best suited to it.
Genetic-algorithm generation of novel matter against the site — scaffold seeding, then evolution under multi-objective fitness. The modality is chosen to fit the biology, not the other way around.
Sources (selection)ChEMBL · ZINC · BindingDB
-
4
The gauntlet
Receives every generated candidate.
Docking and ML affinity, hard selectivity gates against anti-target panels, ADMET and toxicity flags, a synthesizability check, and a novelty test. Most candidates end their run here — by design.
Sources (selection)ChEMBL · BindingDB · SIDER · ToxCast / Tox21
-
5
Lead optimization
Receives the survivors.
Structure-activity analysis and iterative refinement against the same gates — better potency, cleaner selectivity, a route that can actually be run.
-
6
Simulation validation
Receives optimized leads.
The full PK / PD / disease chain — the candidate read out as a drug effect in a patient's biology, not a binding affinity in isolation. The same chain that powers simulations.
-
7
Wet lab & feedback
Receives validated leads with a synthetic route.
Handoff for synthesis and assay — and every result, win or dead end, flows back into the shared foundation, sharpening the next campaign.
Built to break its own candidates
A discovery pipeline that flatters its own output is worse than useless — it manufactures false confidence and burns wet-lab budget proving it wrong. Ours is built to try to break every candidate. Six gates, each of which can end a molecule's run.
Physically sensible
Geometry, strain, and stability that actually hold up. No structures that couldn't exist outside a drawing.
Synthesizable
A real retrosynthetic route. If it can't be made in a lab, it isn't a candidate — it's a picture of one.
Engages the target
Docking and ML affinity against the actual binding site — experimental where structure exists, predicted where it doesn't.
Selective
Hard gates against anti-target panels. Off-target pharmacology and liabilities like hERG are surfaced early, not discovered in the clinic.
Deliverable
Can it reach the tissue at the right exposure? Delivery is a gate, not an afterthought — and in the central nervous system it's the hardest one.
Genuinely novel
New matter, not a rearrangement of what's already known. A candidate that only looks new is a liability dressed as a discovery.
The right kind of molecule — then the journey to it
The therapeutic landscape is no longer small molecules alone. OVIVO reasons across modalities and chooses the one that fits the target and tissue, rather than forcing every problem into the shape of the tool we happen to have.
Delivery is not an afterthought
A molecule that can't reach its target is not a therapy, no matter how exquisitely it binds. Delivery — the right entity, to the right tissue, at the right exposure — is designed for from the start, on equal footing with potency and selectivity. Nowhere does this matter more than the central nervous system, where the barriers that protect the brain are exactly the ones a therapy must cross.
Every candidate carries how much to trust it
The platform won't report false precision. Each generated compound carries an explicit evidence grade, so you know exactly how much weight to place on it. Honest uncertainty is the feature — the alternative is wasted synthesis and failed assays.
Hypothesis
A pharmacophore-based signal only — directional, not a result. Reported as ~50 nM, never 46.7 nM. A place to look, not a place to bet.
Calibrated
Model predictions anchored by docking and checked against held-out data. Enough signal to prioritize a candidate for the wet lab.
Validated
Docking-confirmed and experimentally anchored. And where the platform is blocked — a genuine selectivity barrier, say — it says so. A disclosed barrier beats a confident false positive.
A negative result, rigorously established, is one of the most valuable things a scientific organization can own.
Most of drug discovery is finding out what fails — and the field treats those findings as waste, buried and paid for again by the next group that didn't know. We treat them as a first-class asset. The platform keeps an audited catalogue of what didn't work: mechanisms that failed to transfer, campaigns that hit a genuine wall, calibrations that proved unreliable, and exactly why. When a new proposal resembles a known dead end, the prior failure surfaces — with its evidence — before effort is spent.
We are as proud of the things we've proven won't work as the things we've proven will. Both are knowledge. Both are moat.
Every campaign makes the next one smarter
Discovery doesn't run in a silo. Each campaign forces the platform to confront chemistry and biology it hasn't modeled before; what it learns — a target class's quirks, a calibration earned, a dead end recorded — locks into the shared foundation and sharpens every campaign that follows.
And the loop closes both ways. Discovery feeds the simulation that validates its candidates; the simulation feeds the pathway gaps that begin the next discovery. The boundary moves a little further every time we commit.
More than one way to find a therapy
Generating new matter is one route. The same infrastructure supports several others — each judged through the same chain.
Repurposing
Cross-disease screening finds approved drugs with pathway overlap in a new indication — lower cost, and a head start on safety because the drug has already been in people.
Combinations
For diseases that need multi-target coverage, compound combinations designed with modeled synergy, complementarity, and drug-drug-interaction safety from the start.
Polypharmacology
Single compounds engineered to engage several targets, evaluated for their coverage maps and selectivity across the full anti-target panel.
Retrosynthesis
Route-feasibility assessment, so the candidates that reach prioritization are ones a chemist can actually make — not computational constructs.
Grounded in structure and chemistry
Discovery draws on a curated, growing set of authoritative structural, chemical, and pharmacology databases — a selection below — each retrieved value stored with its provenance.
In a partnership, your sources join it: internal screening data, proprietary libraries, and project-specific chemistry, integrated into a dedicated environment and traced like everything else.
A selection · the full set grows weekly · partner sources integrated per project