To be able to make verifiable decisions, many areas such as AI, biology and medicine rely on precise definitions of terminology that are provided by so-called ontologies. Computer systems can use such ontologies to infer new information from data, thus allowing well-informed decisions in these disciplines. Developing and maintaining ontologies can be challenging, and decisions made using an ontology can be intransparent to end-users. The aim of this project is to develop efficient methods for a problem called concept interpolation, which will improve the situation by supporting a range of tasks from generating definitions to explaining logical innovations with ontologies.

Protein aggregation is believed to be responsible for several degenerative diseases, among which are Alzheimers disease and Parkinsons. The process from a folded, functional protein to disordered, potentially dangerous aggregates containing many proteins, is believed to proceed through an intermediate, partially folded stage. This stage is a critical nucleus consisting of two to ten proteins called an oligomer. Recent findings suggest that these oligomers are more toxic than the mature aggregated fibrils. The short-lived, dynamic nature of these oligomers makes them exceptionally hard to study experimentally. The classical techniques to obtain the structure of a protein, X-ray crystallography and NMR, cannot be applied to this problem. The size of the oligomers along with the unknown structure also makes atomistic simulation studies theoretically and computationally unfeasible. Therefore, we intend to use coarse-grained models to find ensemble characteristics of the oligomers. Initial simulations have been run on Lisa using two separate model: 1) A lattice model that takes hydrogen bonding into account [1]. This model has been used previously to study the aggregation effects for small systems [2]. The model uses a statistical potential for pairwise amino acids, derived using bioinformatics techniques. It also incorporates water interactions derived in a similar fashion. While this model correctly estimates the free energy interactions for residues at ambient temperatures, it does not take the enthalpic and entropic contributions of the solvent into account. Because the relative contributions of entropy and enthalpy to the free energy change with temperature, the free energy of solvation at higher and lower temperatures is estimated incorrectly. Even though we are interested mostly in the rate of aggregation at physiological conditions, experiments are often run at higher temperatures to be able to observe aggregation at a reasonable time scale. To effectively simulate aggregations in these conditions we will extend this model with an effective potential for the hydrophobic effect, developed in our group. These results will be compared with results in a collaborating group in the University of Cambridge. This effective potential uses the statistical potential and combines it with theoretical considerations to approximate the hydrophobic effect well at all relevant temperatures. It has been shown to reproduce thermodynamic temperature dependent experiments quite well for protein folding[3]. This approach will allow us to make predictions for the rate of oligomerization and aggregation at higher and lower temperatures, which can subsequently be compared to experimental data gathered in the Chemistry lab in Cambridge, with whom we have had successful collaborations in the past. They have found that the rate of aggregation depends differently on temperature than the rate of most other reactions. This might be explained by the temperature dependence of the hydrophobe-solvent interactions. If this is the case, this supports the hypothesis that the transition state from folded proteins to aggregates has an exposed hydrophobic surface. 2) An off-lattice model for aggregation with an explicit water model on a lattice. This approach explicitly incorporates water interactions, allowing for an alternative approach to incorporate water entropy. Due to computational and theoretical constraints, this approach is limited to smaller systems and cannot be used to find the trade-off between folding and aggregation. This approach explicitly incorporates water interactions with residues. These approaches should allow us to investigate protein aggregation at two scales: A small system in high detail to investigate the molecular origin of the temperature dependence of the hydrophobic effect, and a large system in lower detail to investigate the role of the hydrophobic effect in the trade-off between protein folding and protein aggregation.

Despite decades of dedicated collaborative efforts, we are still missing key elements to develop the perfect electronic nose. A major obstacle is our limited understanding of how the sense of smell works, and how odorant sensors should be designed for smell detection. The project ODORWISE joins advanced computer simulations together with experimental validation techniques to study how the proteins in our noses can detect and distinguish different chemicals. The outcome will aid the design of new biomimetic sensor units for detecting smells, which will be a significant advancement for early disease diagnosis and surveillance in non-invasive sustainable healthcare.

The recent explosion of experimental data in biological research exposes the complexity of biological systems in unprecedented detail. The wealth of data simultaneously enables and necessitates the construction of models, which allow us to analyse and, ultimately, understand the biological system. Unfortunately, this data does not include the detailed quantitative information that would be needed construct mathematical models of such a system. Formal modelling and analysis techniques that have been developed for distributed computer systems, on the other hand, are starting to be recognised in the biological community as a powerful approach to exploit the available data. We propose to use Petri nets to construct such network models and to apply formal methods for analysis of the model state space. Our Petri net approach will enable the automated construction of models in Systems Biology for cases where existing data is insufficient to construct detailed mathematical models. Petri nets provide a formally defined, intuitive graphical representation that can unambiguously describe biological knowledge. Moreover, the same model can be executed to make testable predictions on the behaviour of the biological system. Model behaviour is sufficiently complex, that the analysis of the state-space will be challenging task and require the development of novel analysis approaches. The automation of parameter calibration together with the graphical formulation will bridge the gap between biology and formal methods, making formal methods directly applicable as an executable biology tool. Specifically, we will engineer Petri net models for the differentiation of blood stem cells into mature blood cells. A better understanding of the blood cell circuitry will help the development of clinical applications such as leukaemia treatment and secure and cheap blood transfusions.

Life sciences have entered the era of big data. The ELIXIR-NL community infrastructure will leverage and develop state-of-the-art technologies to seamlessly bring together very different data -stored at multiple locations-, analytics and high-performance computing. An ‘app store-type’ protected environment will enable scientists to collaborate seamlessly in order to answer complex research questions that could not be answered previously. As a first remit, the infrastructure will be geared to empower personalised medicine and health within the framework of the pan-Dutch Health-RI initiative.