The ISBGroup Blog

We have the pleasure to announce that Maria-Anna Sotiropoulou will be the new contact person for STRATIF-AI.

STRATIF-AI is our flagship project to improve prevention, treatment, and rehabilitation of stroke patients using digital twins and AI. STRATIF-AI has received funding over four years by Horizon Europe, the EU’s key funding programme for research and innovation.

Read more about STRATIF-AI on its homepage

Maria-Anna holds a B.Sc. in Computer Science and Biomedical Informatics and an M.Sc. in Protein Biotechnology. She is currently a Research Assistant at Linköping’s University Department of Biomedical Engineering (IMT), working in the Integrative Systems Biology Group. Her research interests include developing advanced modeling tools for studying intracellular metabolic changes in neurodegenerative diseases, as well as cancer-neuron modeling. She has experience in both experimental and computational research, which complements her interdisciplinary background.

For any administrative inquiries related to STRATIF-AI, from now on, please reach out to both the coordinator, Gunnar Cedersund, and Maria-Anna.

Functional magnetic resonance imaging (fMRI) is one of the most powerful tools in neuroscience, but what does the measured signal really reflect? Traditionally, the fMRI signal has been assumed to mainly reflect excitatory neuron activity. However, recent findings by Nicholas Sundqvist and Henrik Podéus of the ISB Group suggest that inhibitory interneurons play a much more significant role.

Using a model-driven meta-analysis approach, we analysed data across several different studies, revealing that:

• Less than 20% of the fMRI BOLD signal comes from excitatory neurons
• A dominant 50–80% arises from inhibitory interneurons
  

Our model offers a unified mechanistic explanation for why fMRI responses sometimes appear contradictory across experiments. For example, the model provides a plausible mechanism that result in a biphasic vascular response that emerge only under high intensity stimulation.

Taken together, these results point towards a new consensus view: inhibitory interneurons, not excitatory neurons, are the primary drivers of fMRI signals.

We have recently published a manuscript about these findings in the journal Computers in Medicine and Biology – a leading publication in computational biology and medical informatics.

August 2025 was no exception with regard to the ISB Group’s international engagements. After the engaging YoungBMC Symposium at Vrije Universiteit Amsterdam on 26th August where Ralph Monte presented his work “Using Machine Learning Models and Postmortem Metabolomics for Prediction of Ketoacidosis-related Deaths” for fellow young scientists, Ralph, together with Maria-Anna Sotiropoulou headed for the Benelux Metabolomics Days at Leiden University.

The Benelux Metabolomics Days on 27^th and 28^th August brought together researchers from the Netherlands, Belgium and Luxembourg to exchange cutting-edge advances in metabolomics and its biomedical applications. This year’s program emphasized the integration of experimental metabolomics with computational approaches and clinical research, covering diverse areas such as cancer metabolism, microbiome dynamics, exposomics, organ-on-chip technologies, and technological advances in mass spectrometry.

Maria-Anna Sotiropoulou and Ralph Monte were there presenting their work in the poster sessions with titles “Identifying Alzheimer’s Disease (AD)-Associated Metabolic Changes Using ¹³C Metabolic Flux Analysis (MFA)” and “Using postmortem metabolomics to predict ketoacidosis deaths: comparing three machine learning methods” respectively. Maria-Anna presented her work that applied ¹³C MFA to neuronal models of AD to quantitatively map intracellular metabolism. Using ¹³C-labelled Glucose and Glutamine tracers combined with advanced EMU-modeling, preliminary results and flux estimations were presented. In his presentation, Ralph explained how supervised machine learning models can be used to accurately predict ketoacidosis-related deaths. This indicates that machine learning models could be used in forensic settings, e.g., for cause of death determination using the postmortem metabolome.

One of the most impactful talks came from Sarah-Maria Fendt, who presented her group’s research on how liver steatosis worsens colorectal cancer liver metastasis outcomes. Analyzing patients’ data, she showed that liver fat accumulation correlates with the transition from encapsulated metastases (associated with better prognosis) to invasive replacement metastases (linked to poor prognosis). Using mouse models and 3D tumor spheroid cultures, her team demonstrated that excess fatty acids promote tumor aggressiveness through upregulation of proline metabolism, which fuels collagen biosynthesis. This allows cancer cells to build their own extracellular matrix and invade healthy liver tissue.

Alongside this, Marcel Kwiatkowski presented advances in proteo-metabolomics using Metabolic Flux Analysis, highlighting the regulatory interplay between metabolic fluxes and protein modifications. Justin van der Hooft demonstrated how machine learning methods can dramatically expand the reach of mass spectrometry-based metabolomics. He presented computational metabolomics tools that use machine learning to enhance metabolite annotation and analogue searches across complex data sets.

Finally, Alexandra (Sasha) Zhernakova shared insights into microbiome-metabolome dynamics during pregnancy and early life. Her work connects maternal and infant microbiomes with health outcomes, highlighting how early environmental and metabolic exposures shape immune and developmental outcomes.

In August, 4 people from ISB group were in Copenhagen to participate in the “Nordic Ooc/MPS summer workshop”. MPS stands for Microphysiological Systems, and it is another name for experimental models such as organoids, spheroids, organs-on-a-chip, etc. We have been working together with AstraZeneca to develop digital twins for such models for many years, and for this reason we were invited to give two(!) keynote presentations at this event.

The first keynote was held jointly by Oscar Silfvergren from our side, and Liisa Vilén at AstraZeneca. Liisa leads their MPS team, and she and Oscar work tightly regarding modelling, planning of new experiments, etc. They gave a presentation on their collaboration and on the recently submitted project manuscript on multi-scale and multi-species, mechanistic models for MPS and exenatide.

Apart from this, Gunnar Cedersund gave the morning and opening keynote for the second half of the workshop: a hackathon on digital twins for MPS. This second half was a whole-day event to learn the basics of what digital twins are, why they are useful (in medicine, drug development, and for MPS), and also the basics on how to do such digital twin models, via a hands-on computer lab. All participants (38 signed up Ph.D. students and researchers), also had the option of bringing their own data, and getting started on a new project regarding the creation of a model for their own research data and question. Quite a few participants made use of this possibility, and several new collaborations and modelling projects were born during this event.

On the whole, this was a fun, interesting, and useful event arranged by the Nordic network for Organs-on-a-chip, the EU project Open Mind, and their eminent coordinator Jenny Emneus.