MaCSBio strives to perform cutting edge research in the interdisciplinary field of Systems Biology to create a “virtual physiological human”, a set of computational and mathematical models based on biological evidence that will help to understand and predict human systems. Research projects at MaCSBio focus on multi-scale modelling within two complementary research lines tackling areas that are highly relevant for society:
- Systems Medicine of Chronic Diseases - led by Prof. Ilja Arts
- Computational Biology of Neural and Genetic Systems - led by Prof. Elia Formisano
Projects within the “Systems Medicine of Chronic Diseases” research line aim to increase our understanding of chronic diseases and to develop new, biology-based, personalised approaches to prevent and treat them.
With the globally increasing prevalence of obesity and the proportional rise in the aging populations, chronic illnesses such as type 2 diabetes, cardiovascular disease and fatty liver disease emerge as a major challenge for health care. We aim to understand their complexity and large interpersonal heterogeneity, by taking a systems-approach. The focus of our research is on adipose tissue, liver, muscle, and the cardiovascular system. We are modelling whole-body and tissue-specific metabolism and strive to unravel regulatory processes involved in this.
In this research line biological expertise on metabolism, metabolic gene regulation and chronic diseases is combined with a broad spectrum of modelling approaches such as dynamic models, network- and pathway biology, and machine learning and statistical inference.
Dynamic models, to describe and predict changes in metabolism over time
Dynamic models are based on assumptions that are constantly validated against new biological data. They help to predict changes in metabolic processes such as the effect of weight loss on insulin resistance in the adipose tissue of overweight people. At MaCSBio researchers are currently improving a dynamic model of adipose tissue fat uptake and release.
The glycerol part of the model has recently been modified to allow for uptake of glycerol by the adipose tissue, resulting in a significant improvement in the fit of the modified model to measured glycerol data. Several other modifications to the model were also made.
This work highlights the importance of the interplay between computational modelling and data collection. Systems Biology is a constant loop. We use data to generate computational models. Subsequently, data generated by computational models is verified in order to modify and extend these models.
Network- and pathway biology, to analyse, visualize and interpret genome-scale data in their biological context
Many factors play a role in chronic disease onset and progression. While some people develop cardiovascular diseases, diabetes or cancer, others remain relatively healthy. Visualization and (genome-scale) metabolic modelling (GSMM) of molecular pathways and networks in their biological context, can explain these differences and can help to determine the key processes and molecules that are involved.
Such approaches can, for example, be used to find differences in gene expression and metabolic pathways between three different types of patient populations suffering from obesity and presenting either with muscle or liver insulin resistance, or no insulin resistance at all.
Network analysis and pathway biology is especially interesting in the initial exploration and visualisation of complex data. Building larger, integrated biological networks of known molecular pathways allows for the identification of key molecules or processes that can be used in new prediction models.
Machine learning and statistical inference, to generate complex prediction models starting from a hypothesis-free, data-driven perspective
Machine learning and statistical inference go hand in hand in analysing large biological datasets to, for example, identify and subsequently predict the effect of diet in a heterogeneous group of overweight people.
A major challenge in systems medicine is to link different types of models into a single complex system, “The virtual physiological human”, representing the human body, which can be effectively used in both research and clinic to predict e.g. disease progression and treatment effects.
These data-driven approaches are able to jointly model associations between all factors on every scale, from cells to organs and organ systems. Moreover, machine learning techniques enable us to automatically build accurate prediction models, which can be easily validated and updated once new data becomes available.
The research line “Computational Biology of Neural and Genetic Systems” aims to increase our understanding of brain function by integrating data and results from investigations across different neuroscience disciplines.
We investigate the mechanisms underlying learning, in particular perceptual learning. This is highly relevant for society, given the increasing necessity of life-long learning, and the mounting human and financial cost of impaired learning and related brain pathology. We are taking a network approach to understand and model structural and functional interactions among elements such as neurons or neuronal populations as well as genes or proteins.
This research program requires expertise in systems neuroscience and specifically in mathematical and computational modelling, computational neuroimaging and neurogenetics.
Mathematical and computational modelling
At MaCSBio we use different computational models and modelling approaches to bridge several neuroscientific research domains. These include single neuron- and neuronal population models, neuroenergetics, neuronal synchrony, and multisensory processing, and ultimately generate integrated multi-scale models of the brain. The Kuramoto model, for example, was recently extended with new parameters to study the effect of learning on structural and functional changes in a network of neurons.
Results are promising, however a detailed characterisation of all factors that affect the network dynamics and structure is required prior to applying the model to explain empirical data.
At MaCSBio, neuroscientists are being trained to become multidisciplinary researchers. They need integrated multi-scale models to generate specific predictions from one spatial scale to the next scale, and vice-versa.
Computational neuroimaging, to study the neurobiological basis of learning in healthy subjects as well as subjects with increased or decreased learning abilities
Non-invasive imaging methods, such as MRI, allow investigating the brain on a larger spatial scale, on the level of populations of neurons. They provide information on e.g. brain anatomy, interactions between different brain regions and regional brain activity over time.
Learning can be investigated by asking a subject to perform easy visual and/or auditory tasks before and after learning. Non-invasive imaging methods show the changes in the brain induced by learning.
Computational models help me as a neuroscientist by linking my observations with non-invasive imaging to underlying neurogenetic processes. This is a necessary step in developing new theories and to increase our understanding of perception and learning.
Neurogenetics, to identify candidate genes regulating learning
Advances in computational biology approaches have made it possible to identify genes that play a key role in crucial cellular processes. The process of mitochondrial replication is in particular interesting for learning. Mitochondria are the energy-factories of our cells and are specifically important in tissues with a high-energy demand, such as the brain.
Defects in mitochondrial replication have been identified in patients with neurodegenerative disorders, including learning disabilities. Recent computational analyses have successfully identified candidate genes for an essential role in the replication of mitochondrial DNA.
As molecular biologist at MaCSBio, it is my job to validate if the hypotheses resulting from computational models are biologically correct. My research proves if these candidate genes are indeed involved in mitochondrial DNA replication and whether mutations in the identified genes are found in patients with severe mitochondrial dysfunction. Moreover, we explore the role of these genes in learning and memory.
Cross-talk between research lines
While the two research lines have high scientific and societal impact on their own, MaCSBio aims to explore the cross-links between them. Given the high prevalence of chronic diseases and brain pathologies, this may have an enormous public health impact. Using a systems-approach, we will generate relevant breakthroughs beyond current research. A relation between the research lines is apparent since increasing evidence becomes available associating type 2 diabetes and cardiovascular diseases with (mild) cognitive impairment and increased risk of dementia. Disruptions in energy metabolism might be at the basis of this observation, since muscles, liver and brain are all tissues with high-energy demand. Finding the biological causes of these associations is an important focus of current biomedical research.
Resulting models will predict, for example, how dysregulated glucose metabolism affects organs directly involved in maintaining normal glucose levels as well as neuronal synapses in the brain, and to diagnose whether a person with diabetes will develop cognitive impairment or not.