/ via embl/
Building on his bioinformatics expertise, EMBL alumnus Moritz Gerstung is developing the next generation of AI tools for cancer research
In 2021, Moritz Gerstung left EMBL-EBI to become the Division Head for Artificial Intelligence and Cancer Evolution at the German Cancer Research Centre DKFZ, and a professor at the University of Heidelberg, Germany. His lab uses machine learning to understand the growth and development of tumours.
Before that, Moritz Gerstung was a Research Group Leader at EMBL-EBI (2015–2021), focusing on understanding the genetic mutations caused by cancer. During his time at EMBL-EBI, Gerstung and his group revealed fascinating new insights, including:
Below, Gerstung talks about how his time at EMBL-EBI propelled him into his current role and supports his work exploring how AI can supercharge cancer research.
How do you use AI in your work?
Cancer tumours are ecosystems of billions of mutant and normal cells. Charting and modelling all this information requires sophisticated learning and AI tools. Our lab develops such tools, and carries out large-scale data analyses to understand cancer evolution.
When I started at EMBL-EBI in 2015, I was focusing on bioinformatics analysis of cancer genomes. During my time there the artificial intelligence (AI) capabilities and expertise in my group grew, which enabled me to move to this new role at DKFZ, where my group predominantly focuses on AI tools.
Does your research focus on a specific type of cancer?
While I was at EMBL-EBI, we focused on leukaemias for practical reasons. The tissue is relatively easily accessible and people had already assembled large cohorts that were amenable to systematic examination.
We showed, for example, that acute myeloid leukaemia – initially considered to be one disease – was in fact eleven different diseases. This partially explains differences in clinical outcomes. This is the kind of situation where precision medicine opens up new, more personalised treatment avenues instead of always relying on the sledgehammer of chemotherapy.
When I transitioned to Heidelberg, my focus grew on investigating brain tumours because there is a lot of expertise here in this field.
Many new AI developments are coming from the private sector. Are you concerned that open access to these tools might be compromised by the scale and cost of developing, training and using AI algorithms?
It’s an important question, but I think one critical element is the accessibility of data, and this is where EMBL-EBI is one of the international ‘lighthouses’ because it makes available so much of the world’s biological data. AI systems, including AlphaFold, were only possible thanks to EMBL-EBI’s data resources. Making sure that these continue to exist and grow is key to developing AI algorithms in public and private settings.
Are you worried that AI might be misused?
We need to provide regulatory frameworks to prevent abuse of biomedical information to discriminate against individuals. There are already some checks and balances in place, but we need to make sure they’re still up for the task.
I’m convinced AI will make our lives easier because it will help information flow more easily and help to make us more productive. AI enables us to break new ground and gain new insights.
What do you think is next for AI in molecular biology?
I think AI will bring many incremental improvements. Over time, we’re likely to see an acceleration in the improvement of our understanding because we can investigate things faster, and avoid more failures. I think AI will become part of our day-to-day toolkit as scientists.
What did you enjoy most about working at EMBL-EBI?
Life on the Wellcome Genome Campus – where EMBL-EBI is located – is very special because it’s a very close-knit community. You’re always bumping into colleagues in the coffee queue or walking from one building to another, and that creates a great atmosphere of scientific exchange and developing new ideas, which I think is very special.
Are there any particular people or moments who influenced your career?
It all started with Janet Thornton, who was the director of EMBL-EBI at the time and who recruited me. I remember having fascinating conversations as part of the interview process, and thinking that EMBL-EBI is a place with really nice, intelligent and trustworthy people.
Working at EMBL-EBI was also an opportunity to get access to large volumes of electronic health records from the UK Biobank, Denmark’s national health registries, to gain insights into how diseases develop over a lifetime and how pre-conditions and lifestyle influence health and disease risk.
Crucially, my time at EMBL-EBI helped me develop fruitful long-term collaborations, including one that recently resulted in the development of an AI model that can forecast disease risk for more than 1000 medical conditions.
What was your proudest achievement at EMBL?
Using earlier generations of AI tools to understand the microscopic characteristics of tumours and for the first time, link them to the molecular characteristics. This revealed how closely intertwined the molecular and the histological worlds are. Until then, this had not been possible simply because of the lack of appropriate tools.
To find out more, watch this short video interview with Moritz Gerstung below or listen to the full-length interview podcast, conducted by EMBL Alumnus Angus Lamond, and edited by the EMBL Alumni Relations team.
