Team: Linda Douw en Matthan Caan
Neurological and oncological diseases affect the body and brain at multiple scales: from cells, to organs, to the whole body. A major problem in biocomputation is combining data from these different scales of measurement, particularly when wanting to create new effective treatments. The end goal of this project is to develop computational techniques that bridge the microscale and macroscale in neurological and oncological diseases.
Subproject 1 will focus on the fact that many treatments for oncological and neurological diseases are ‘shooting in the dark’: monitoring of cellular pathology, like Alzheimer pathology or cell division in cancer, is almost impossible without invasive procedures such as biopsies. This subproject will make it possible to use macroscale, non-invasive Magnetic Resonance Imaging (MRI) to visualise microscale pathology. Using Artificial Intelligence (AI), we will create “MRI microscope models”, which are able to pick up on cellular pathology based on macroscale/MRI
images. These MRI microscope models exploit deep learning techniques in a new way, making use of the unique multiscale data available within Adore. In neurology, we will first focus on iron deposition as a cellular pathology relevant to Alzheimer’s Disease (AD). In oncology, we will exploit the MRI microscope models to better pick up on and predict accelerating cell division, particularly in neuro-oncology. Ultimately, these MRI microscope models have the potential to be used across a range of oncological and neurological diseases, where effective treatment is hampered by the impossibility of visualising and monitoring cellular pathology.
Subproject 2 targets the issue of multiscale complexity of neurological and oncological disease: these pathologies span from genes to proteins to cells to organs to the whole body. Moreover, these scales are not independent and influence each other, making it rather difficult to pick the most relevant treatment target to best care for individual patients. This subproject will develop methods that can synergize between these different data scales, while also taking crossscale interactions into account. For this, we will use “multilayer network theory”, as the rich knowledge on how such networks work will help us tame the complexity of the different data sources and interactions. As a starting point, we will use multimodal brain imaging in combination with one other data type to build such a multiscale, multilayer network toolbox. In neurology, we will focus on MS (integrating neurofilament light and brain networks) and AD (integrating proteomic profiles and brain networks), while our oncological focus will be on brain cancer (integrating tumour genotypes and brain networks). At the end of the project, we aim to be better able to tailor treatment of cognitive deficits, for instance through non-invasive brain stimulation targeted at
particular brain regions, if this proves the most relevant layer.
Together, these tools, shared amongst and outside of Adore, will allow fundamental and clinical researchers to get a grip on the massive amounts of multiscale data associated with neurological and oncological disease, and exploit them in the most optimal way to either inform further mechanistic work or do clinical or translational studies. Although we focus on particular applications of both the MRI microscope models (iron deposition and cancer progression) and the multilayer network toolbox (cognitive decline) for feasibility, they can be seen as beginnings of a moonshot research line that will allow biocomputation to bridge the micro and macroscales.
