Tuesday 7 November 2017, 09:00-17:00
Eindhoven University of Technology
Grand Cafe ‘The Black Box’, TU/e campus.
Chantal Tax, PhD, Cardiff University Brain Research Imaging Centre, UK:
‘Unravelling the brain’s connections with MRI’
Show abstract and bio
Chantal Tax (1988) was born in Weert, The Netherlands. She completed her Biomedical Engineering master thesis in the group of professor Bart ter Haar Romeny in collaboration with epilepsy center Kempenhaeghe, cum laude. The focus of this project was to improve the reconstruction of the optic radiation for epilepsy surgery, using functional MRI and diffusion MRI tractography. After graduation in 2012 she acquired a PhD degree at the PROVIDI Lab, part of the Image Sciences Institute, also cum laude. Under the supervision of Dr. Alexander Leemans, she focused on improving the analysis of diffusion MRI data. In Nov 2014 she received the Marina van Damme Grant at TU/e enabling a six months scientific project at the Laboratory of Mathematics in Imaging at Brigham and Women’s Hospital and Harvard Medical School, Boston, MA. In April 2017 she was awarded an NWO Rubicon Grant, from which she is currently doing a two year research project on the super-MRI with strong gradients of the Cardiff University Brain Research Imaging Centre in the UK. Chantal’s current research focuses on exploiting the ultra-strong gradients in multi-modal MRI to make a comprehensive assessment of the white matter microstructure in the living human brain.
For decades, neuroscience has focused on unravelling the function of brain areas but only relatively recently on communication between these areas through white matter pathways. The advent of diffusion MRI (dMRI) and diffusion tensor MRI (DT-MRI), which can probe structures at much smaller scales than the imaging resolution through the random motion of water molecules, has undoubtedly increased the interest in studying white matter in the living brain.
With dMRI, both the geometry and microstructure of white matter pathways can be investigated. This presentation will give an overview of the current state-of-the-art in measuring meaningful geometrical and microstructural features, and will highlight the opportunities and pitfalls in obtaining the most complete picture of the white matter to date. The crucial role of using ultra-strong gradients in this process will be discussed.
Prof. Dick de Ridder, Bioinformatics Group of the Plant Sciences Group, Wageningen University & Research Centre:
‘Machine learning to understand and engineer biomolecular interactions’
Show abstract and bio
Prof. Dick de Ridder leads the Bioinformatics Group, of the Plant Sciences Group at Wageningen University & Research Centre. He was a postdoc, assistant and associate professor in the Delft Bioinformatics Lab in 2003-2013, and in 2013 he moved to Wageningen. With a background is in pattern recognition / machine learning, his research goal is to develop adaptive algorithms and models for molecular biology, primarily based on high-throughput measurement data and available prior knowledge.
His presentation will focus on his research theme: “Developing algorithms that enable us to integrate biological data with existing biological knowledge to build predictive models, which in turn help biologists. E.g. using data on protein sequence and function to predict protein-protein interactions, which helps to understand biological networks. Our biological application is currently mostly in plant sciences, where we focus on crops relevant to agriculture, particularly vegetables and fruits. We want to understand how their taste and appearance are determined at the genetic and molecular level. Very similar questions are relevant to animal, food and nutrition sciences.”
Bioinformaticians develop databases, tools and algorithms to store, process and analyze the massive amounts of data generated in modern molecular biology. As many molecular processes in living cells and organisms are still too poorly understood to construct physicochemical models, there is ample room for machine learning approaches. Indeed, sequence-based machine learning has been applied in bioinformatics almost since the start of the discipline. Nowadays learning algorithms are routinely used on a variety of data types, to predict properties of molecules, such as gene function or protein location, and to predict interactions between molecules.
Increasingly, machine learning algorithms are also used in systems biology, to model (parts of) of cellular processes. The most interesting aspect of such models is not so much the prediction but the actual combination of features used, which may yield insight into the underlying molecular mechanisms. The complexity and performance of such models critically depends on the accuracy and amount of measurement data available. I will present a few examples of recent work in which we model processes that are poorly understood at various level of detail: translation, internal ribosome entry, and protein secretion.
At the same time, molecular techniques and laboratory technologies are becoming available that allow us to synthesize large numbers of molecules (genes) and test their activity in living cells. The combination of high-throughput automated experimentation and machine learning is set to revolutionize biology: rather than modeling biology and testing hypotheses on individual parameters, we can set up large experiments aiming to train full models and test their performance. In such cases measurement data can even explicitly be produced to develop a machine-learning based model. I will present some initial results and ongoing projects that take this approach to optimize protein secretion, gene expression and flavour compound production.
Registration is free for members, and 15 EUR for non-members. You can register here.