Multiscale brain modelling: toward brain digital twins
Full Professor of Physiology
Deputy leader of Dept of Brain and Behavioral Sciences, University of Pavia
Deputy leader of Human Brain Project, WP1 (EU)
Leader of the Brain Connectivity Unit at IRCCS C. Mondino, Pavia
President of the EBRAINS Italian Community
Addressing the multiscale brain organization is fundamental not only to understand its inherent mechanisms of function but also to answer neuropathological questions and promote the development of new technologies for AI and health. While relevant advances have been made on the experimental front – encompassing genetics, molecular biology, cell physiology and brain imaging – recent developments in informatics and big data have opened a new scenario, in which multiscale computational models can be used to simulate brain functions and to foster a range of technological applications. Multiscale brain modelling is an emerging technological sector. In principle, it should be possible to model neurons and synapses in detail and then connect them into large neuronal assemblies to explain the relationship between microscopic phenomena, large-scale brain functions, and behavior. More difficult is to infer neuronal functions from ensemble measurements like those currently obtained with MRI, EEG, MEG or PET. In this presentation, I will consider theories and strategies for combining bottom-up models, generated from principles of neuronal biophysics, with top-down models, based on ensemble representations of network activity and on functional principles. Modelling the relationship between microscopic phenomena and large-scale brain functions could allow us to predict how a drug that binds specific receptors modifies local and distributed circuit activity or how genetic alterations of membrane ionic channels or receptors reverberate up to brain functions and dynamics. This, in turn, would allow us to identify potential targets for pharmacological and physical therapy, e.g., through electrical or magnetic stimulation of specific circuits, or for precision surgery. Clearly, these applications open new perspectives toward personalized and precision medicine, for example generating brain digital twins. These can be intended as personalized copies of a subject’s brain that can be used to simulate specific functionalities anticipating the consequences of, e.g., neurorehabilitation or surgical intervention. Multiscale brain modeling has also breakthrough potential in information technologies and AI. Spiking neural networks can be transformed in hardware to generate neuromorphic computers and be embedded inside closed-loop controllers to generate new computational architectures and autonomous robots. In conclusion, multiscale brain modelling is not just fundamental to understand brain functioning but also to promote digital technologies for society and health in ways that remain to be worked out and exploited in full.