And most of what each programme learns never reaches the next.
The advantage isn’t a better model. It’s better judgement.
Drug discovery is driven by decisions, not predictions. Dalton keeps every model, decision and experiment, the dead ends included, in one place with its full context. Your scientists set direction, agents do the computational work, and every experiment strengthens the system and informs the next decision. What worked, what failed and why stays with your team, so the next programme starts where the last one finished.
Discovery cycles that learn.
Dalton reads your project data, compound registration, ELN, even your reports and documents. Its agents design and prioritise the next molecules. Your labs or your CRO partners make and test, the results flow straight back. Design, make, test, learn, orchestrated end to end, with each round sharper than the last.
Agentic AI. Knowledge. Tools.
Seven layers. The tools do the computational work. Context & knowledge makes the system yours. Agentic AI runs the loop across all of it, and the digital twin shows what that was worth.
Hundreds of tools at the base, all of it inside your own isolated tenant. The layers above are what create lasting value and deliver real impact.
What would Dalton have done with your campaign?
The twin replays your real campaign on its real timeline and asks how a different decision policy would have done. Same molecules, same measured data. Measured, not asserted.
Purpose-built for every modality.
The same platform underneath. On top, the science each modality actually needs.
Affinity, selectivity, ADMET and physchem from as few as 5–20 compounds, the regime of early discovery. No retraining per task, so cycles stay fast.
LLM-driven generation that works the way a medicinal chemist does, weighing physchem, scaffold and synthetic constraints to propose novel, makeable molecules.
Docking, free-energy and quantum-mechanical methods, run as durable workflows and applied when needed, where cheaper scoring cannot decide.
Experimental effort focused on the compounds that will teach you the most, so each round of synthesis generates the maximum information.
Public and internal sequence sources, including the full Observed Antibody Space, annotated and clustered so the patterns across your library surface.
Structure prediction, inverse folding and protein language models that generate and optimise sequences, fine-tuned from as few as 20 examples.
Physics and machine-learning scoring combined to rank candidates on affinity, developability and structural fit, with active learning cutting the compute to a fraction of brute force.
Sequence-diverse antibodies that keep the same epitope binding mode, spreading immunogenicity risk and widening IP coverage.
Agents propose. Your scientists approve.
Agents read every file you upload. Nothing becomes project truth until a person releases it.
Discovery knowledge
you own.
Dalton runs in your own isolated tenant, behind your own single sign-on. Your data trains only your models, and the judgement it builds stays with you, not the vendor.
Built by the people who built AI discovery at AstraZeneca, Exscientia and Oxford University.
Announcements, partnerships and platform news.
DaltonTx launches its drug discovery platform, Dalton, and announces Sygnature Discovery collaboration
Bonito Biosciences & DaltonTx announce AI-driven collaboration to advance precision delivery for oligotherapeutics
DaltonTx raises £4 million seed financing to build the intelligence backbone of drug discovery
Better decisions. Better molecules.
Faster.
For pharma, biotech and CRO teams running discovery. Tell us what you’re working on and we’ll set up a working session with the team.
368 Gray's Inn Road
London
WC1X 8BB