Leading the charge of AI in life sciences
Interview with Stein Aerts, Director of VIB.AI: the Center for AI and Computational Biology
October 2, 2024
Little more than a year ago, VIB’s 10th research center, VIB.AI: Center for AI & Computational Biology was launched. Heading this new endeavor is Prof. Stein Aerts, Scientific Director of the new center focusing on applications of artificial intelligence and computational biology. We sat down with Stein to reflect on the past year and to discuss what the future holds for this exciting new field of research.
Hi Stein! You have been heading the VIB.AI Center for Computational Biology since June of last year. How has that been going so far?
It has been really exciting to launch VIB.AI. We established strong connections to multiple research departments at KU Leuven and UGent to which our group leaders will be associated. What’s really great, is that both computer science and biomedical research groups participate. Three key areas we focused on are obviously the international recruitment of new VIB.AI group leaders, the connection within VIB to the other nine centers through 12 co-affiliated PIs with strong computational and AI expertise, and the setup of our GPU compute infrastructure at the Data Core and VSC, the Flemish Super Computer center.
And how has that been going? Any new colleagues to announce?
We just announced our first Group Leader, Joana Pereira. Her new team will focus on proteome-wide structural bioinformatics. We also have our new call open, for two group leaders in Ghent and two in Leuven.
Apply for our open Group Leader and Professor Positions: AI in Biology in GhentApply for our open Group Leader and Professor Positions: AI in Biology in LeuvenAside from recruiting, what else has been going on?
Another essential element that we have been fine-tuning over the past year, is our strategic plan. VIB.AI is a bit different from other research centers at VIB since we are the first center that is embedded in more than one university: we have structural ties with KU Leuven and Ghent University, and we also have co-affiliated PIs at VUB and UAntwerp.
We assembled our first Scientific Advisory Board, which will visit us in November. Together with this team of world experts, we’ll discuss our vision on AI research for modeling biological systems. We hope to concentrate on in-depth modeling of complex biological systems, using both deep representational learning and mechanistic models based on first principles.
We have set up two important components to facilitate this. Firstly, on the computational side we invested in top-notch compute infrastructure so our PIs and co-affiliated PIs have all the resources available that they would need to conduct their research. For instance, we were able to acquire a cluster of H100 GPUs from Nvidia on which our scientists will be able to train their models. We are hosting the VIB Data Core which will be invaluable for the research that we will be conducting in the future. Furthermore, we are setting up a Machine Learning expertise unit to provide AI & software engineering support to our researchers. Secondly, on the experimental and technologalical side, we are launching a Wet-lab and Technology unit on which our core PIs can rely for their data generation and experimental validations. I find this an exciting setup, reciprocal to the other centers that usually have wet-labs and a bioinformatics support unit. We’ll have all-dry labs and a wet-lab support unit instead.
That all sounds very exciting indeed! Can you shed some more light on how the co-affiliation of PIs at VIB.AI works?
Ah yes, so the co-affiliation principle starts from the idea that we want to foster a community of PIs from other VIB research centers, each with their own specific expertise in computational biology, to address interdisciplinary biological questions. The beauty of this, is that it works in both directions: VIB.AI can rely on all kinds of biotechnology expertise, be it in plant systems biology or oncology research, whereas the other research centers have a direct link to our own activities and resources. It truly fosters collaboration and ensures that we keep each other updated on developments in our respective fields.
The co-affiliated PIs and the VIB.AI PIs form one community with joint seminar series, scientific meetings, faculty meetings, PhD calls, shared GPU and software resources, collaborative science projects, and a lot more.
And the same goes for partners outside of VIB too! We connect with computer science and computational biology experts at the Flemish universities. We notice that AI in biology is one of the key topics in the life sciences across Flanders and we want to be at the forefront of those conversations. We host monthly seminars for around 30 research labs where researchers from Flanders and Brussels can connect and share experiences. In the future, we hope to do the same at an international level as well and connect with other leadings institutes in Europe.
Over the past two years, AI has really claimed its place in the limelight. Do you see a future without AI?
I really don’t think so. AI is much more than just a hype. I predict we will see more and more integration of AI in all types of research within life sciences.
Personally, I think the real value of AI in biology lies in its ability to model complex biological processes, and reveal mechanistic insights into molecular processes, cells, tissues, organisms, ecosystems, and diseases. Within VIB we have access to state-of-the-art technologies that produce large amounts of data that we can use to train our models. With AI we have the power to integrate various types of data and come to predictions and insights that we would not have been able to anticipate by traditional hypothesis driven research. AI is just an incredible efficiency gain in that regard.
How do you see AI itself evolving in the future?
We are seeing new developments in machine learning and AI every day, so it’s hard to say where it will go. An interesting futuristic model will enable crossing the various layers from genotype-to-phenotype, with multi-scale causal inferences from the DNA sequence over protein interactions, cell-cell communication, cell function and tissue function. Imagine for example an AI model that predicts purely from the genome sequence the complete spatial transcriptome atlas of an organism. Another exciting prospect is the use of current ‘black-box’ models to derive ‘white-box’ mechanistic models that are fully interpretable.
And how does VIB approach this evolving landscape?
We believe that AI sits at the intersection of technology and biology. With VIB.AI we are ambitious to bring together a large team of AI experts and computer scientists, while being fully integrated in the VIB life science environment. In that sense, VIB was one of the early life science institutes to adopt “AI in Biology” as a key research area, whereas now we start seeing this evolution taking place at other institutes too. We also commit to improving on the technology itself, allowing us to make the difference and be pioneers in the field.
What kind of impact are we talking about?
There are a lot of possibilities in, for instance, drug discovery. VIB has a unique strength and track record in translating research into actual value for society, be it in the form of start-ups or by collaborating with industry. We believe we can contribute to societal impact with our research at VIB.AI, working towards the creation of new therapies for patients or solutions for agriculture. From personalized medicine and digital twin, to antibody design, and synthetic biology applications in cell & gene therapies. The possibilities are endless.
I’m curious to see how agentic AI will further influence our brainstorms in the future.
Right now, we should capitalize on the power of AI and data-driven research, to discover new insights and applications in biomedicine, and to solve diseases that we haven’t been able to solve so far.
Thank you, Stein! It’s clear to see the possibilities with AI are vast, and we’re just getting started. I’m excited to see where the future of VIB.AI takes us.
