From London to Lausanne: how Isomorphic Labs is rewriting drug discovery

Sergei Yakneen became the Chief Technology Officer of Isomorphic Labs in 2022. Thomas Kern / SWI swissinfo.ch

Google DeepMind spin-off Isomorphic Labs is building an AI drug design engine that it believes can “solve” all diseases. We spoke to its Chief Technology Officer in Lausanne about Switzerland’s key role, AI hype, and what it will take to cure cancer.

In 2013, Sergei Yakneen was on the fast track to a brilliant tech career, managing a team of software engineers at Amazon in Toronto, Canada, when his life took a dramatic turn. His mother died of pancreatic cancer aged just 54.

Yakneen grew up in the 1980s surrounded by doctors and scientists in Krasnoyarsk, a major industrial city in central Siberia, which had a reputation as a land of exile for opponents of the Russian regime.

Yet he never felt the calling to become a doctor like his mother, an oncologist at the city hospital. “I was more interested in computers,” said Yakneen, who would spend hours assembling computers and writing computer programmes as a kid.

But watching the doctors in his family search for a treatment for his mother’s cancer took him in a different direction.

“I felt helpless in my tech career,” said Yakneen, who was 33 at the time. “I couldn’t stay in the tech world in good conscience. I kept thinking to myself – in the future, if one of my kids is confronted with the same disease, what will have been my contribution to the solution?”

He quit his job at Amazon to join the Ontario Institute for Cancer Research. There he used his tech know-how to analyse DNA sequencing data from thousands of cancer patients to understand how genetic mutations shape the onset and progression of cancer.

Almost a decade later, in 2022, Yakneen became the Lausanne-based Chief Technology Officer and part of the founding team at one of the most talked-about companies in the AI drug discovery field – Isomorphic Labs.

About Sergei Yakneen

Sergei Yakneen grew up in Krasnoyarsk, Siberia, and immigrated to Toronto, Canada aged 14, after the fall of the Soviet Union. He spent the next 21 years in Canada, where he studied computer science and mathematics and was first exposed to neural networks and machine learning. He worked at several different companies, including e-commerce giant Amazon, launching its first Canadian software engineering organisation.

He then became a research assistant at the Ontario Cancer Research Institute and helped to lead the technical working group of the Pan Cancer Analysis of Whole Genomes Project – the world’s largest cancer data analysis initiative.  

He went on to do a PhD in computational biology and cancer genomics in Heidelberg, Germany, where he wrote algorithms for analysing genomic data at population scale. This landed him a job at SOPHiA GENETICS, a Swiss healthcare technology company that creates AI platforms to help clinicians and researchers interpret complex genetic and clinical data. Yakneen joined Isomorphic Labs as Chief Technology Officer in 2022, shortly after the company’s founding in November the previous year.

The London-based company was spun out of Google’s AI lab DeepMind in 2021 with the mission to “solve all diseases with AI”. Its founder and CEO, Demis Hassabis, co-won the Nobel Prize in Chemistry in 2024 with his colleague John Jumper for developing a groundbreaking AI model, AlphaFold. Together they’d solved a problem chemists had wrestled with for over 50 years: predicting the three-dimensional structure of a protein.

Most drugs work by binding to precise points in a protein, which is only possible by knowing its 3D shape. Identifying this shape was a slow and expensive process that could take as long as five years, until AlphaFold came along and did it in minutes.

“It was clear right away how transformative this model was,” said Yakneen, who was working at SOPHiA GENETICS when the second version of AlphaFold (there are now three) was released in late 2020. “It could take an entire PhD to decipher the structure of a single protein. With AlphaFold 2, you could punch in the amino-acid sequence and get the 3D structure prediction at nearly the same level of accuracy.”

It would be hard to find a pharmaceutical company or biotechnology lab in the world that isn’t using AlphaFold or an AlphaFold-inspired structure-prediction model at some stage of their research.

On its own, AlphaFold can’t cure diseases, but that was never the goal, said Yakneen. “We are building a whole suite of AI models that all together, in concert, are our drug design engine that we believe will create medicines faster, more economically, and at much higher success rates.”

>> Find out more about AlphaFold and how it solved a 50-year-old mystery

This is exactly what drew him to the company. “The mission is not just about going after one particular disease or one sliver of the drug discovery problem, but to build general models that will help us eventually solve all diseases,” he said. This includes diseases long thought undruggable.

The “grand challenge”, he said, is getting to the point where our AI models can design a variety of molecules with very specific properties – almost like tailor-made medicine for a specific person and disease.

Isomorphic is now a key player in an increasingly competitive race to revolutionise drug development with AI.

Pharma giants Novartis and Eli Lilly, developer of the blockbuster GLP-1 weight-loss drug Zepbound, have already partnered with Isomorphic to discover new drug candidates in deals that could be worth billions. In January, it signed a deal with US healthcare giant Johnson & Johnson to tackle “difficult to drug” disease targets across treatment modalities. Isomorphic is also developing its own drug candidate portfolio focused on oncology and immunology.

The third version of AlphaFold was jointly released by Isomorphic Labs and Google Deepmind in 2024, allowing researchers to model the structures and interactions of all of life’s molecules. The company has since developed further breakthroughs that tackle the full spectrum of complexity in drug design.

The Lausanne connection

As Chief Technology Officer, Yakneen is in charge of building a team to develop the technology behind Isomorphic’s “drug design engine” and suggested Lausanne as the place to do it alongside its London headquarters.

Yakneen fell in love with the city after relocating there in 2019 to work at SOPHiA GENETICS, where he oversaw the development and operation of a global AI-based molecular diagnostics platform.

Perched above Lake Geneva, Switzerland’s fourth-largest city is home to several universities, including the Swiss Federal Institute of Technology Lausanne (EPFL) – one of Europe’s top science and engineering universities, on par with the US’s prestigious Massachusetts Institute of Technology.

“I was immediately struck by both the beauty of this region but also how entrepreneurial and sophisticated the local tech and AI scene was,” he said. “I thought anyone would want to be here if they’re not here already.”

In May 2023External link, Yakneen opened Isomorphic’s site in Lausanne at the EPFL’s Innovation Park, allowing the company to tap into the area’s “vibrant scientific and technical community,” he said at the time. But it quickly outgrew the location and now has a permanent home in the former industrial district of Flon.

So far, he has recruited around 30 people, largely data engineers and machine learning researchers but hopes to start taking on more staff with pharmaceutical and biotech backgrounds. The company has over 300 employees globally and is adding 25 new positions in Lausanne or London, according to its website.

The hiring spree has been made possible by $600 million (CHF460 million) in external funding secured in March 2025 from US venture capital firm Thrive Capital, with participation from GV, formerly known as Google Ventures, and Google’s parent company, Alphabet. Isomorphic Labs also received $45 million upfront from Eli Lilly and $37.5 million from Basel-based Novartis in the first collaborations announced in 2024.

From dream to reality

There are high hopes and expectations that AI can overcome the time-intensive, complex, and risky business of drug development. Experts estimateExternal link that AI could reduce the cost and time for drug development by up to 50%.

Today it takes at least a decade and $2.5 billion to develop a drug, according to some studiesExternal link. Some 90% of drugs don’t even make it to market because they fail clinical trials when tested for safety and efficacy in humans.

The use of AI in the development and discovery of drugs is still in its early days, but in some quarters there is scepticismExternal link that AI can really deliver the benefits and breakthroughs its proponents claim are on the horizon. There are a few AI-discovered candidates in late-stage clinical trials but not one has been approved by regulators.

Isomorphic has faced its own setbacks. Hassabis told an audience at the World Economic ForumExternal link in January 2025 that the company hoped to have AI-designed drugs in clinical trials by the end of the year. But on January 20, at this year’s forum, he acknowledgedExternal link that timeframe has been pushed back to end-2026.

Data quality and availability remains a key bottleneck in AI-driven drug development. “With machine learning, we’re deploying these learning algorithms on top of vast datasets to be able to make predictions,” said Yakneen. “But we still need to generate even more data.”

He recently joined the scientific advisory board of the UK Biobank, the world’s largest longitudinal study of over 500,000 UK volunteers that includes detailed genetics, imaging, clinical and lifestyle data. “These resources are exactly the types of large and high-quality healthcare research datasets that will power the next generation of AI breakthroughs,” he said.

There has also been pushback over the decision to limit access to the software code of AlphaFold 3. It was eventually made available for non-commercial use six months after its release, but this contrasts with the release of the full source code of AlphaFold2.

Yakneen defended this decision. “There isn’t a hard and fast rule that says all of this should be open source,” he said. “There needs to be a balance, and we thought about this very carefully with AlphaFold 3 and decided to publish the paper and the method so everybody could benefit from it.”

As a commercial entity, he added, “you need to keep certain things private so that you can maintain your competitive edge and a commercial endeavour”.

Despite the challenges, Isomorphic has stuck to its mission statement to one day “solve all diseases” with the help of AI, a mission Yakneen acknowledges will be a long march.

Yakneen fell in love with Lausanne when he moved to university town in 2019. Thomas Kern / SWI swissinfo.ch

“We’re developing these new technologies that have huge promise, but in the end, we need to make sure that these medicines are safe,” he said. “Nothing will be gained by racing through and then putting something on the market that isn’t going to help patients.” 

He remains committed to finding a cure for the cancer that took his mother’s life. “Pancreatic cancer is really, really difficult. There’s been only marginal improvements to treatments,” said Yakneen. “It remains unsolved, but the capabilities we are developing give me real hope.”

Edited by Nerys Avery/vm/gw