Drug-discovery company Evotec and biotech Odyssey Therapeutics have formed a partnership to use artificial intelligence to find new treatments for autoimmune and inflammatory diseases.
Announced August 7, 2026, the collaboration combines Evotec’s data-driven discovery platform — AI data science, advanced screening, compound libraries and disease-modeling tools including stem-cell-based and multi-omics approaches — with Odyssey’s expertise in the biology of immune-driven diseases.
What a hit series is, and why payment attaches to it
Evotec may receive milestone payments when it delivers validated hit series — promising early drug candidates — for the targeted diseases. Financial terms were not disclosed.
The term is specific. A hit is a compound showing the desired activity against a target in a screening assay. A hit series is a group of structurally related compounds with that activity, which is considerably more valuable than a single molecule: it demonstrates the activity comes from a definable chemical scaffold rather than an experimental artefact, and it gives chemists variants to work from when optimising for potency, selectivity and drug-like properties.
Validated means the activity has been confirmed in follow-up testing — a necessary filter, since screening produces many apparent hits that prove to be assay interference rather than real binding.
Why the milestone is set there
Paying on hit series rather than on clinical progress reflects a realistic division of responsibility, and it is informative about what each side believes it controls.
Getting from a target to a validated hit series is a defined technical problem: build the right assay, screen enough chemical matter, confirm what emerges. It is precisely the stage where computation, automation and large compound libraries help most, and it is the stage Evotec sells.
Everything after — optimisation, toxicology, clinical trials — depends on biology no platform controls, and most failure occurs there. A discovery-services company sensibly declines to be paid on outcomes it cannot influence.
Where AI genuinely helps
“Drug discovery increasingly depends on the ability to integrate deep disease biology with advanced experimental and computational approaches,” said Evotec chief scientific officer Cord Dohrmann.
The concrete contributions are narrower than the label suggests. Computational methods predict which compounds are worth synthesising and testing, reducing the physical screening burden. They identify structural patterns across chemical libraries that human chemists would not extract. And they integrate multi-omics data — genomics, transcriptomics, proteomics across many samples — where the dimensionality exceeds manual analysis.
What they do not do is determine whether a target actually drives disease in patients. That is where most drug development fails, and it is not a computational problem.
Why autoimmune disease is a hard target class
Immune-driven diseases present a specific difficulty that shapes what the partnership is attempting.
The immune system is necessary. Suppressing it treats autoimmunity and creates susceptibility to infection and cancer, and the broad immunosuppressants used historically illustrate the trade-off plainly.
Modern approaches aim for selectivity — blocking a specific cytokine or signalling pathway central to one disease while leaving general immune function intact. That is why the biology expertise Odyssey brings matters as much as the discovery capability: choosing which pathway to target is the decision that determines whether the eventual drug is useful, and it is upstream of anything a screening platform does.
Part of a pattern
The deal is one of many pairing AI-and-data platforms with disease-focused biotechs, as the industry bets that blending computation with biology can shorten the slow, failure-prone early stage of discovery.
Evotec separately noted that Niagen Bioscience picked it to advance a small-molecule candidate for rare genetic diseases and accelerated aging.
What disease modelling contributes
The stem-cell-based disease-modeling tools mentioned in the collaboration address a problem that sits upstream of screening chemistry.
Drug discovery has historically tested compounds in immortalised cell lines — convenient, cheap, endlessly divisible, and frequently unrepresentative of the human tissue they stand in for. Many have accumulated genetic changes over decades of culture and behave unlike the cells in a patient.
Induced pluripotent stem cells offer an alternative. Ordinary cells from a donor are reprogrammed to an embryonic-like state, then differentiated into the cell type of interest — immune cells, in this case — carrying that donor’s genetics.
For immune-driven disease that is particularly useful, because susceptibility is strongly genetic and much of it lies in genes governing immune regulation. Testing a compound in cells from patients who actually have the disease is a closer approximation to the eventual clinical question than testing it in a generic cell line — and it is the kind of capability that justifies a partnership rather than an internal effort.
Multiple concurrent partnerships are the point of the model. A services platform monetises the same infrastructure repeatedly across partners, spreading the cost of capability that no single small biotech could justify building. Business news, not investment advice.