Impressions from the ECCB2026

Last Friday Beatriz and Bernhard returned from the European Conference on Computational Biology, ECCB. It was held at the Geneva “Centre International de Conférences de Genève (CICG)”, where in 1985 Reagan and Gorbachev met to end the Cold War. So the plenary sessions took place in a massive auditorium with doubled rows of chairs. Apparently delegates at the type of international meetings the center was originally designed for are paired with assistants.

Of course, the scientists at the ECCB had no assistants whispering into their ears, but the existential dread of the Cold War years was back when Jeremy Farrar, Chief Scientist at the World Health Organization, called 2026 an inflection point in history. And not a good one. He went on to urge everyone to work towards a more equitable world, as our current problems wouldn’t be sorted out by others. In spite of these problems, Farrar said, we are living in a golden scientific moment, just think of the state of Computational Biology in 2001, the year of the first ECCB.

Concentrating on the science, Beatriz and Bernhard said these were their favorite parts of ECCB2026:

Beatriz: This conference had three highlights for me. Two of them were the keynotes by Anna-Sapfo Malaspinas on “Tracing migration from ancient DNA” and by Aleksandra Walczak on “Learning about immune response from large-scale data”. Malaspinas explained how ancient DNA can help us better understand the diseases of the past and how they spread going into the present. While the main topic of the talk was the bacterium Treponema pallidum subsp. pallidum, which causes syphilis, it was interesting to also learn about the other three treponematoses (yaws, bejel and pinta), which are caused by other Treponema species and strains. It would be intriguing to see whether we can detect marker regions that can be used to distinguish these four strains from each other.

In the talk by Aleksandra Walczak it was impressive to hear about the diversity of antigens that we all carry inside ourselves, and how these can both serve as an individual’s fingerprint but also as a reflection of our immune system’s history.

The third highlight for me was the focus session entitled “From data to discovery” with Nick Loman, Angie Hinriches and Theo Sanderson. It gave a fascinating insight into how to handle and visualize the incredible amounts of data that large-scale epidemiological studies can produce. It will certainly be useful to take a look at the software they used, UShER, for “Ultrafast Sample Placement on Existing Trees”, to see how it might help with our own research questions.

Bernhard: For me the most interesting part was the workshop on large language models on the last day. Ever since the excitement about AlphaFold, I have been wondering whether large language models could be used for classifying bacterial genomes. I still don’t know, but the hands-on coding in the workshop using a dedicated repo helped me get started. While working through the example Jupyter notebooks we were given, I was struck by how demanding it is to just fine tune LLMs, let alone train them from scratch. Our fine-tuning runs lasted several minutes on the GPU machines of our Google-Colab accounts. Fat chance doing this on a laptop. Still, the only way to demystify a technology is to play with it, and the workshop was an excellent starting point for playing with large language models.