The Netherlands eScience Center is excited to announce the eight winning proposals for the ASDI 2020 Call. The ASDI call solicits research projects with innovative but challenging domain research questions that require eScience technologies and software. In this way, the eScience Center seeks to support research endeavours in which the application of eScience tools and methodologies is relatively underdeveloped.
The winning proposals are classified in four broad discipline areas:
- Environment & Sustainability
- Humanities & Social Sciences
- Life Sciences & eHealth
- Physics & Beyond.
Each of the winning projects receives a grant consisting of funds and in-kind support by Research Software Engineers from the eScience Center.
1. Environment & Sustainability
Computing societal dynamics of climate change adaptation in cities
Dr. Debraj Roy (University of Amsterdam)
This project addresses the gap by employing state-of-the-art computational and data analytics methods to advance the science of climate change adaptation (CCA). To achieve this goal, the project team will rely on their disciplinary knowledge, their small-scale computational spatial agent-based model (SABM), a unique micro-level data on CCA from four countries, and the specific skills that the research software engineers at the eScience Center offer. The project will innovate by (1) revealing hidden patterns in the high-dimensional survey, going beyond the standard regression analysis in social CCA studies; (2) scaling up the SABM to systematically explore the factors reinforcing poverty traps as climate change intensifies; and (3) building web-based decision-support to examine the cumulative effect of individual CCA. The project team will develop reusable software that is accessible to a wide range of users and permits experts to trace socio-economic resilience to climate change.
Accelerating Process Understanding for Ecosystem Functioning under Extreme Climates with Physics-Aware Machine Learning (EcoExtreML)
Dr. Yijian Zeng (University of Twente)
This project will couple the vegetation photosynthesis model (SCOPE) with the soil moisture model (STEMMUS, considering dynamic root growth), synergized with Earth-Observation data, to understand how the water-carbon dynamics of an ecosystem vary with environmental and climate stress. Currently, a hurdle in applying STEMMUS-SCOPE globally is its expensive computational cost. As a first step, both STEMMUS and SCOPE will be exposed to the Basic-Model-Interface, which will be subsequently coupled using Python-Modeling-Toolkit, serving as first level acceleration. Second, a physics-aware machine learning emulator, based on a limited number of STEMMUS-SCOPE runs, will be developed. Furthermore, to address the ‘data-gap’ issue of satellite reflectance products (i.e., revisit-time (5–27days) and cloudy condition), OpenDA will be deployed to assimilate multiscale/multi-sensor data to generate spatiotemporally continuous information on ecosystem functioning. This project will open a variety of new opportunities to observe earth in order to retrieve higher-level products like root-zone-soil-moisture and belowground-carbon-allocation, in addition to land-atmosphere gas exchanges.
2. Humanities & Social Sciences
Impact and Fiction (ImpFic)
Dr. Peter Boot (Huygens Institute for the History of the Netherlands)
This project aims to find out which properties of fictional texts have an impact on readers. The types of impact will include affective responses to narrative and style, as well as reflection. Earlier research often handled all readers as essentially similar and targeted a single measure of success (popularity or sales). They also tried to predict success based on features that are hard to interpret from a literary point of view (such as word frequencies). In this research project, different groups of readers will be distinguished. The project team will use a large corpus of recent Dutch novels (10,921 texts), a large corpus of online book reviews (> 472,000) and a large collection of book lists (> 37,400) created by users on book-oriented social media sites. In the reviews, the team will measure different types of impact. Based on the book lists, the team will then cluster readers by their preferred type of reading. For the texts themselves, the team will define new metrics for key textual properties that are potentially partly responsible for the impact a book has on its readers. These metrics will include parameters referring to the novel’s narrative, writing style and mood.
Navigating Stories in Times of Transition: The COVID-19 Pandemic as a Use Case
Prof. dr. Gerben J. Westerhof (University of Twente)
This project aims to develop eScience technologies to advance multidisciplinary approaches around digital storytelling. When societal transitions happen, people respond by telling stories that create continuity and change. In a digital age, storytelling has become even more abundant, rich in variety, and accelerating in pace. The current COVID-19 pandemic allows us to study storytelling while transitions are happening. A multidisciplinary team will analyze how the varieties of stories evolve over time and across different media. eScience challenges include (1) integrating existing Natural Language Processing tools to extract and preprocess stories from existing and growing collections; (2) further developing a Digital Story Grammar methodology to study narrative components in their mutual relations; (3) building a Story Navigator that uses visualization tools, while also allowing for human interpretation; and (4) integrating the tools into the Orange platform to allow re-use by students and researchers. The team will cooperate with national infrastructures, develop online tutorials, and organize a hands-on workshop to facilitate dissemination and long-term use. Interdisciplinary narrative studies will be enriched with necessary tools to analyze storytelling in the digital age as it unfolds in real time across different media and platforms. In this way the research can potentially support sustainable policy making in times of transition.
3. Life Sciences & eHealth
Transformer-based deep learning for next generation mass spectrometry-based phosphoproteomics
Prof. dr. Connie R. Jiménez (Amsterdam UMC)
This project aims to apply state-of-the-art AI technologies to a unique collection of phosphoproteomics data. When successful, this project will change current practice in DIA-MS analysis, and catalyze both cancer-signaling research and biomarker and target discovery enterprises to ultimately improve cancer diagnosis and treatment. Multiple oncogenes encode protein kinases that are involved in aberrant cellular signaling by phosphorylation, which is a target of many FDA-approved drugs. Phosphoproteomics by mass spectrometry (MS) provides a global view of cellular protein phosphorylation, making it highly relevant to cancer research. Important for clinical proteomics, a recent MS advancement called dataindependent acquisition (DIA) allows for the high-throughput generation of quantitative proteomics data in a more comprehensive fashion. Standard DIA requires a pre-existent library of high-quality MS spectra for molecules to be considered. A challenge for DIA application in phosphoproteomics is the lack of a spectral library for the plethora of possible phosphoprotein modifications. So far, only project-specific libraries are created, which have limited depth, and require considerable instrument time.
Digital biomarkers for Parkinson’s disease: gaining real-life insights from wearable sensors
Dr. Twan M. van Laarhoven (Radboud University)
Parkinson’s disease is the fastest growing neurological disorder worldwide, with an expected doubling of patients to 12 million by 2040. Disease-modifying treatments are being developed, but testing their efficacy is hampered by the lack of objective biomarkers. Unobtrusive wearable sensors now allow us to objectively capture and continuously monitor how patients function in their natural environment. However, reference datasets are scarce, as are models to extract relevant insights from the raw sensor signals. To build new models to monitor Parkinson’s disease symptoms, the project team will exploit recent developments in unsupervised deep learning, combined with others methods that are not fully supervised. The results will be integrated in a modular toolbox for clinical researchers to efficiently pre-process sensor data and extract digital biomarkers for Parkinson’s disease progression. Such open-source tools have the potential to revolutionize the design of clinical trials, and create new opportunities for telemedicine.
4. Physics & Beyond.
High-throughput GPU computing for New Physics searches with electrons in LHCb
Dr. Roel J.M. Aaij (Nikhef)
In recent years, the LHCb experiment has found tantalizing hints of new quantum effects in very rare processes in which so-called B-particles decay to muons or electrons. Although combinations of several of such measurements seem to indicate a consistent picture (recently referred to as the ‘flavour anomalies’), no clear single observation of a so-called “New Physics” phenomenon has been made. This project aims to perform such a measurements by introducing a new computing scheme, in particular allowing the improved detection of electrons and neutral (uncharged) particles. The upgraded LHCb detector is currently being installed and is scheduled to start taking data in the autumn of 2021. Given current expectations, and with the proposed computing scheme, this project has the unique potential to make a first and ground-breaking discovery of new fundamental particles. A key component of the upgraded LHCb detector is the processing of an unprecedented data rate of 4 TB/s in software on a heterogeneous farm of general purpose graphics processors (GPGPUs) and CPU servers. This research proposes to extend the GPGPU-based data filter with advanced capabilities for electron reconstruction and classification to enable the discovery of new physics.
Unravelling Proton Structure with Hyperoptimised Machine Learning
Dr. Juan Rojo (VU University Amsterdam)
This project aims to tackle long-standing puzzles in our understanding of strong interactions, from the origin of the proton spin to the strange content of nucleons. The key to achieving this will be the first-ever universal analysis of nucleon structures from the simultaneous determination of the momentum and spin distributions of quarks and gluons and their fragmentation into hadrons. This research will combine an extensive experimental dataset and cutting-edge theory calculations within a machine learning framework where neural networks parametrise underlying physical laws while minimizing ad-hoc model assumptions. Given that computing resources represent a major hurdle in this ambitious research program, it will be crucial to optimize the model training by exploiting GPUs. Furthermore, the exploration of the resulting complex parameter space demands an algorithmic strategy to determine model hyperparameters such as network architectures.
The team at the eScience Center congratulates all winners, and thanks all participants for the variety and diversity of strong submissions. Together with our extended research community, we look forward to the results of these projects in due course.