Picture a pharma team that has just watched its Phase II trial fail. Years of work and tens of millions spent, and the drug didn't move the needle on the chosen disease. Five competitors are working on the same target. The obvious move is to shelve it.
But here’s the thing: the molecule might work - just in a different disease, or for a different patient group. Of the drugs that eventually get approved, roughly 60% end up treating another condition better than the one they were first tested for. The problem is that the only way to find out is to run another trial - and at $20m and three years a time, most teams never get the chance.
This is the problem Atlas Bio is solving, and when Simon first walked us through how he sees a world where human biology is simulated, it immediately clicked.
Models like AlphaFold changed how we understand protein structure, and a new generation – Chai-2, IsoDDE, BoltzGen – are pushing toward designing biologics directly in silico. Beyond individual models, the rise of “AI scientists” - systems that can synthesise thousands of papers, design experiments, and iterate on results without human input - will compress R&D timelines in ways that weren’t conceivable two years ago.
But the clinic itself hasn’t changed: a Phase II trial still takes two to three years, costs $20m or more, and roughly 70% of drugs still fail at this stage.
This is what Atlas are building for.
The Atlas Bio team
The team has built a foundation model trained on roughly 2 million RNA readouts from human tissue - a signal that shows how genes switch on and off in the body. Because the model learns from human biology rather than from drug treatment data, it can simulate how any drug with a known mechanism will behave across a virtual cohort of patients, without needing sensitive patient data. What would take years in the clinic becomes a 12-week simulation. And since the trial is virtual, you can do things that are impossible in a real one: test several treatments on the same patient profile, or run a drug across all diseases at once to see which patients actually benefit.
That is exactly what happened with a large pharma client. After running an Atlas simulation across their pipeline, they identified a promising new indication for an asset they had written off - and are now pursuing it. This is not a one-off: Atlas is already under contract with a top-10 global pharma, in active conversations with more than a dozen of the world’s largest drug makers, and every prospective prediction made on real trial outcomes has proven correct.
Simon has been an outlier from early on: as a teenager he won Jugend forscht, Germany’s national youth science competition, and placed third out of 8,000 students in his state’s Mathematical Olympiad. He went on to study at KIT, EPFL and Stanford and worked on self-driving as early as 2015. He also has been living with type-1 diabetes, and so has benefited throughout his life from good and rapidly improving treatments.
This triggered his passion for the space - and before writing a single line of code for Atlas, he spent a year having 300 conversations with people across drug development to understand where things really break. He is incredibly curious, gets everyone around him excited, and is relentless: he once ran 140km across the Swiss Alps in one go, simply because nobody had done it before.
His co-founder Sebastian spent over a decade in deep learning, including LLM research at Google. And Dominik, their CSO, spent over a decade in post-doctoral research and translational medicine, where he used an early version of this exact approach to save a Phase II trial. He joined Atlas because it had already been his passion project for years.
More broadly, we think foundation models are now enabling a class of prediction problems that simply weren’t tractable two years back - across physics, materials science, and biology alike. Atlas are going after probably the highest-stakes of all: whether a drug will work in humans, before first patient enrollment.
Simon, Seb, Cam and Kathi after signing the termsheet
With Atlas more good drugs can reach the patients who need them, sooner. We're proud to be leading Atlas Bio's $11m Seed round, and can't wait to support Simon, Sebastian, Dominik and the team on this journey!
