Dutch Public Reaction on Governmental COVID-19 Measures and Announcements
Public sentiment (the opinion, attitude or feeling that the public expresses) has been always attracting the attention of government, as it directly influences the implementation of policies. In the current epidemic situation, understanding the opinion of general public timely becomes even more important. However, the ‘staying-at-home’ policy makes face-to-face interactions and interviews challenging. Meanwhile, about 2.8 million users in the Netherlands use Twitter to share their opinions, making it a valuable platform for tracking and analyzing public sentiment. To understand the variation of Dutch public sentiment during the COVID-19 outbreak period, this project will analyze real-time Twitter data using machine learning and natural language processing approaches. Data collection will be based on COVID-19 related keywords and users. Its aim will be to provide a cost-effective and efficient way to access public reactions in a timely manner. For instance, instead of waiting for physical behaviours (like taking a walk outside) of people, their sentiment and intended behaviour could already be gleaned from Twitter data.
Research team: Dr. Shihan Wang, Dr. Marijn Schraagen, Prof. Mehdi Dastani (Utrecht University)
eScience Research Engineer: Dr. Erik Tjong Kim Sang
Diagnostic imaging with computed tomography (CT) and chest X-ray are proving increasingly important in detecting and assessing disease severity of COVID-19. To aid in the clear communication between radiologists and clinicians the Radiological Society of the Netherlands (NVvR) has proposed a standardised reporting system, CO-RADS, for assessing the suspicion and severity of COVID-19 in a CT scan. In this project, an existing platform, grand-challenge.org (with elements of the the EYRA benchmark platform, where suitable), will be furthered to enable the development and deployment of machine learning algorithms for automated scoring of CT scans using the CO-RADS system.
Research team: James Meakin, Paul Gerke, Mike Overkamp and Miriam Groeneveld, Prof. Bram van Ginneken (Radboud UMC)
eScience Research Engineers: Maarten van Meersbergen, Pushpanjali Pawar, Jesus Garcia González
Diagnostic information and data on occurrence of cardiovascular complications in COVID-19 patients is rapidly growing but is distributed over different clinical locations. In order to provide the most accurate insights about the relation between cardiovascular history and related complications in COVID-19 patients, statistical analyses and machine learning models need to be kept up to date in real time. This will not be possible by continuously collecting data manually from different locations. The FAIR Data for Capacity project will build FAIR data stations and automatic data extraction pipelines for defined sets of clinical data as part of a distributed learning infrastructure. This will provide insight in the incidence of cardiovascular complications in patients with COVID-19, and the vulnerability and clinical course of COVID-19 in patients with an underlying cardiovascular disease.
Research Team: Dr. Andre Dekker (Maastricht University, Personal Health Train – PHT), Dr. Rick van Nuland (Lygature, HealthRI), Prof. Folkert Asselbergs (UMC Utrecht, Dutch Cardiovascular Alliance – DCVA), Dr. Mira Staphorst (Hartstichting, DCVA)
eScience Research Engineers: Dr. Lars Ridder, Djura Smits, MSc
Real Time National Policy Adjustment and Evaluation on the Basis of a Computational Model for COVID19
The current COVID-19 pandemic presents an unprecedented challenge for policy makers. Although the major consequences from the uninhibited spread of COVID-19 virus in Western European countries have abated due to far reaching social distancing measures, these measures carry enormous economic and social costs. Furthermore, basic epidemiological mechanics dictate that some form of containment policy will be necessary for the foreseeable future in order to prevent a recurrent outbreak and keep the impact of COVID-19 manageable. The challenge then is to design public policy interventions informed by epidemiological models. However, these models suffer from what has been termed in other fields the curse of locality: while the basic biology of the virus is the same everywhere, the outcomes will differ according to the local circumstances: the host population in each country is different, societal norms and customs vary and spatial patterns governing movement of people in their daily lives differ. This means that Dutch policy must be informed by a model that is tailored to circumstances in the Netherlands. In this project, work will continue on developing an epidemiological model that can be used to inform public health interventions and is specifically tailored to circumstances in the Netherlands.
Research team: Prof, Martin Bootsma, Prof. Marc Bonten (UMC Utrecht), Prof. Jason Frank (UU), Prof. Mirjam Kretzschmar (UMC Utrecht, RIVM)
eScience Research Engineers: Dr. Inti Pelupessy, Dr. Ben van Werkhoven, Dr. Rena Bakhshi, Lourens Veen, MSc
The microbes in our bodies are fundamental to our health. At the molecular level, many of their interactions with human tissues are mediated by microbial specialized metabolites.
While metabolomics provides a powerful technique to profile these, most microbial molecules have unknown structures; hence, over 95% of detected masses cannot be functionally interpreted or linked to their producers. This currently thwarts efforts to understand important diseased states of our microbiome.
Many innovative computational workflows have recently been designed to predict molecular (sub)structures from genomic or metabolomic data; however, these efforts have remained largely unconnected. Integrating these data will make it possible to complement partial information provided by each field to yield much better functional predictions.
Moreover, it will connect vital information from both data types: while metabolomics informs about in vivo relevance, genomics informs about biological origin. Here, we propose to design a novel algorithm to connect molecular substructures identified in tandem mass-spectrometric data to sets of genes within biosynthetic gene clusters (BGCs) detected in (meta)genomic data. Subsequently, we will integrate this algorithm with our previous methods for metabolome (spectral networking, substructure detection) and genome analysis (BGC identification and clustering) in one comprehensive eScience workflow.
Finally, we will demonstrate its potential by identifying molecules prominent during periods of relapse in a longitudinal study of inflammatory bowel disease (IBD) and connecting them to their producers. Ultimately, our workflow will illuminate the vast unknown metabolic space within the human microbial metabolome, and greatly advance our understanding of molecular mechanisms of health and disease.
The FAIR guiding principles for data management and stewardship (FAIR = Findable, Accessible, Interoperable, Re-usable) have received significant attention, but little is known about how scientific protocols and workflows can be aligned with these principles.
Here, we propose to develop the FAIR Workbench that will enable researchers to explore, consume, and produce FAIR data in a reliable and efficient manner, to publish and reuse (computational) workflows, and to define and share scientific protocols as workflow templates. Such technology is urgently needed to address emerging concerns about the non-reproducibility of scientific research.
We focus our attention on different types of workflows, including computational drug repositioning to illustrate fully computational workflows and related systematic reviews to illustrate mixed (manual/computational) workflows. We explore the development of FAIR-powered workflows to overcome existing impediments to reproducible research, including poorly published data, incomplete workflow descriptions, limited ability to perform meta-analyses, and an overall lack of reproducibility.
We will demonstrate our technology in our use case of finding new drugs and targets for cardiovascular diseases, such as heart disease and stroke. As workflows lie at the heart of data science research, our work has broad applicability beyond the Life Science top sector.
Automated Medical Image Analysis (MIA) has the potential to truly innovate clinical practice by offering solutions to routine, yet key tasks, such as segmentation (i.e., delineating organs). Especially with recent advances in machine learning (ML), in particular in Deep Neural Networks (DNNs) that have led to an explosive growth of successful MIA studies reported in academic literature, the time appears right for such innovations to find widespread real-world uptake.
Yet, labor-intensive manual performance of these tasks is still often daily clinical practice. In this proposal, we integrate DNNs with other state-of-the-art computational intelligence techniques, in particular evolutionary algorithms (EAs), to overcome 2 key obstacles in moving toward widespread clinical uptake of computationally intelligent MIA techniques: 1) observer variation in the definition of a ground truth, and 2) image quality variation due to different acquisition protocols and scanners at different institutes.
In particular, we design and develop efficient-computing-compatible implementations of mixtures of DNNs, the results of which can be fused with results learned from other data sets (i.e., from different institutes).
To maintain sufficient focus while doing so, we consider an elementary, but key MIA task: segmentation. Moreover, by means of an application in radiotherapy treatment planning, in collaboration with the Academic Medical Center in Amsterdam, we validate the newly developed technology on real-world patient data within the runtime of the proposed project.
The project aims to identify genetic variants associated with sleep patterns, and to perform Mendelian randomisation studies to identify the downstream causal consequences of disturbed sleep patterns on metabolic diseases such as obesity and type 2 diatbetes. UK Biobank offers a large and high quality dataset to perform above analyses. Over a hundred thousand individuals wore a wearable movement sensor (accelerometer) on their wrist for the duration of one-week (24/7). Previously, eScience Engineer Vincent van Hees developed and published software to analyse this kind of accelerometer data. This was successfully used in preliminary analyses of UK Biobank. The sleep detection functionality of his software is designed to combine sleep diary information with accelerometer information. However, in UKBiobank there is no sleep diary data. Vincent implemented a possible solution for this but did not publish an article on it. In the project we will evaluate the currently implemented algorithm to detect human sleep from wearable accelerometer data without the aid of sleep diaries. Next, we will attempt to develop a better method, e.g. with machine learning, and release as update to the existing open source software.
Interactions between biomolecules control all cellular processes. Understanding those interactions requires adding a three dimensional structural dimension. Next to experimental structural biology techniques, this can be done by docking, a complementary and high-throughput computational method allowing to model complexes from their known components.
A challenge in docking is scoring – the identification of correct (near-native) models from a large pool of docked models – due to our still limited knowledge of interaction rules. We will tackle this challenge by training deep networks (dNNs) to learn complex interaction patterns from the huge amount of experimental data in the Protein Data Bank (a valuable source of information not yet fully exploited). Our innovative strategy is to treat this problem as a 3D image classification problem: The interfaces of docked models will be represented as 3D images and dNNs will be trained to classify whether they are near-native or not. Unlike other machine learning techniques, dNNs are now able to learn from millions of data without reaching a performance plateau quickly, which is computationally tractable by harvesting GPU and Hadoop technologies.
The resulting scoring function, DeepRank, will markedly enhance our capability to reliably model biomolecular complexes, assisting the scientific community to gain insights into macromolecular aspects of life. It will be implemented in our HADDOCK modelling platform and freely distributed through GitHub and eStep repositories, ensuring a wide dissemination. The impact will be broad since 3D image-based dNNs have applications in many other domains, such as medical diagnoses (MRI), cryo-electron microscopy and computer vision.
Co-applicant: Dr. Li Xue (Utrecht University)
Image by: NIH Image Gallery
Winner of the Young eScientist Award 2015
Most people with active epilepsy live in rural areas in low and middle income countries. Despite the availability of cheap and save drugs, more than 80% is not on treatment due to lack of required diagnosis. Development of diagnostic methods that are available in rural setting and that do not rely on medical specialists will greatly help to reduce this large treatment gap. Recently, there have been promising results in using computers to diagnose epilepsy based on standard electroencephalography (EEG) recordings.
This path-finding project will bridge the gap between electroencephalography prediction modeling and epilepsy diagnosis in rural areas of Sub-Saharan Africa, where the burden of epilepsy is highest. EEG will be acquired in community services with wireless, affordable consumer-grade headsets. Diagnostic EEG prediction models will be developed with machine learning and signal processing algorithms. Models will be incorporated in a telemedicine web-portal to aide future diagnoses and reduce the significant burden of epilepsy.
Image: Sadasiv Swain – Distributing anti-epileptic drugs in Jalda, India