Mechanistic Modelling in Pharmacology, Physiology, and Neuroscience

Professor, Director of SciLifeLab

The rapid development of new imaging technologies allows unprecedented insights into the molecular machinery inside living cells and organisms. For the first time, light and electron microscopy have molecular sensitivity and resolving power in situ, and, if used together, can connect structural detail with molecular dynamics of the whole cell. Aided by machine learning driven image analysis powered by open sharing of image data, this provides unprecedented opportunities for new insights into the molecular mechanisms that drive life’s core functions at the scale of the cell.

I will present the progress we have made to study one of life’s most fundamental functions, cell division, by mapping the dynamic protein network, assembly of individual protein complexes and genome re-folding that drive it. Our work has studied cell division in human cancer cells and early mammalian embryos using advanced cross-scale imaging methods, including light-sheet, quantitative fluorescence correlation spectroscopy (FCS)-calibrated, super-resolution and correlative light and electron microscopy. The quantitative integrated molecular data that these new technologies deliver, allow us to better understand how the molecular machinery functions in space and time to ensure faithful cell division and prevent the errors that underlie congenital disease, infertility and cancer.

SciLifeLab fellow, KTH Royal Institute of Technology

Focus TBA

DDLS Fellow, Linköping University

Learning molecular variability from cryo-EM data relies on strict manual control of correlated variables, or the use of general ML frameworks that have to learn the correlative variation in addition to estimating it. This presents a complexity trade-off that is prone to bias in either modality, preventing maximal use of current datasets. Here we present an approach that seeks to parameterize 3D cryo-EM reconstructions in maximally interpretable ways, based on general causative mechanisms, that can then be recast to provide explicit learning objectives through an explicit model. This seeks to minimize the bias incurred by specifying ad-hoc mechanistic models, and efficiently regularize general learning frameworks.

Professor Stockholm University, Linköping University and KTH Royal Institute of Technology, Director of National Academic Infrastructure for Super­computing in Sweden (NAISS)

Focus TBA

Professor, Linköping University, WASP Graduate School Director

Focus TBA

DDLS fellow, The Computational Microscopy for Cell Biology (CMCB) laboratory, Stockholm University

Focus TBA

Focus TBA

Professor, Charité University Medicine Berlin

This talk explores the integration of mechanistic modeling, multiscale physiological data, and artificial intelligence to construct personalized “Digital Brain Twins” for neuroscientific research and clinical applications. It addresses the computational, methodological, and ethical challenges in ensuring these virtual replicas remain dynamic, precise, and robustly validated across spatial and temporal scales. Ultimately, the presentation highlights frameworks for building trustworthy in silico models capable of predicting disease progression, evaluating targeted pharmacological interventions, and advancing personalized medicine in neurology and neuroscience.

Associate Professor, Linköping University

Focus TBA

Professor, Lund University

One of the hottest scientific topics today is how does the brain actually work and, in the era of artificial intelligence perhaps even more, what is ‘intelligence’ really? Here we will address these questions to illustrate the path towards arriving at a general conceptualization of the brain circuitry operation that emerges from the fundamental dynamics of neuron interactions, i.e. from neurophysiological observations. In this ambition, there is also a need to explain the resulting roles fulfilled by the different structural subsystems of the nervous system. Fundamental differences with current AI systems is one outcome. Consequences relative to the wider literature are; (i) the function of any one neuron alter depending on context; (ii) in this mode of operation predictive neural activity emerges as a trivial by-product without any assumptions of architectural specializations. This underscores the need for integrated neurophysiological work, as the broader field is still far away from understanding the physiological infrastructure that underlies the emerging computational architecture of the brain.

Vice President and Head of Neuroscience at Lundbeck & Affiliate Professor at University of Copenhagen

How AI and data-driven approaches are currently being integrated into neuroscience drug discovery, and how these approaches can be combined with mechanistic understanding across different biological scales – from molecular and cellular mechanisms to circuits, systems and ultimately clinical phenotypes. A perspective on where the main opportunities, but also the current limitations, of AI and mechanistic modeling in pharmaceutical research.

Senior Researcher, Chalmers University of Technology

Metabolic pathway maps are familiar, but they cannot tell you what happens when a gene is lost or a nutrient runs short. Genome-scale metabolic models turn the same biochemistry into something computable and testable against real perturbations. I will introduce Human-GEM, the community model of human metabolism, show a few applications across tissues, organs and disease, and suggest that models of this kind become more valuable as omics data accumulate, by giving those measurements a mechanistic structure to be interpreted against.

Professor, KTH Royal Institute of Technology

Focus TBA

Focus TBA