EGI COVID-19 support activities: an update

Giuseppe La Rocca and Gergely Sipos update us on the EGI initiative to support COVID-19 research projects

In alignment with many other organisations and project initiatives, EGI and Open Science Grid decided to join forces and provide specialised technical support, simulation tools and data intensive resources to promote international cooperation in research to tackle COVID-19. The call for COVID-19 related projects was opened in April 2020.

Six months later, EGI and OSG have successfully started to support the following COVID-19 initiatives:

  • VINI – multi-drug multi-target docking service for COVID-19. Draško Tomić and his research team from the Rudjer Boskovic Institute, Croatia, have accessed EGI cloud resources to investigate possible new drug combinations, their inhibition on SARS-CoV-2 spike glycoprotein and perform fast virtual screening of novel drug candidates.
  • Animal genomics for a “One Health” perspective in the COVID-19 pandemic era (AnGen1H). Luca Fontanesi and his team from the University of Bologna – Department of Agricultural and Food Sciences are using the resources of the INFN-Bari provider to apply large scale genomic data analyses in pets and livestock species. Overall, the project aims to mine their genomes for potential variants that might confer resistance to SARS-CoV-2 and other coronaviruses that could cause devastating economic effects by disrupting livestock production chains. A total of about 20TB of data have been used for the pilot setup. A preprint of a research article describing the first results obtained in pigs is now available online.
  • Contact Tracing: A Machine Learning Approach. Kallol Roy and his team from the Institute of Computer Science, University of Tartu, Estonia, have started to use EGI resources to develop a machine learning-assisted COVID-19 contact tracing application. The app can be installed in smartphones, tablets, is privacy enabled and it does not share patients private data with third party vendors. The causal relations for human-to-human transmission of COVID-19  is extremely complicated to model mathematically. So the team is using machine learning to learn the human-to-human transmission automatically from COVID-19 data. The app predicts the probability of getting infected beforehand, based on an individual’s profile (e.g. age, places visited in the last few days etc). The app can alert people and also tells them the probability measure of their infections. The COVID-19 data set is large and high-dimensional and so the team is deploying EGI cloud hardware for both training and inference. The computing model and the datasets to train the application are now tested in the EGI cloud Infrastructure. This work will be using the Jupyter Notebooks and Cloud Compute resources provided by EGI. The training of the network is expected to take 2 weeks using High-Throughput Compute resources provided by EGI.

In addition, EGI is also supporting Alexandre Bonvin and his team from the Utrecht University to port HADDOCK in the EGI cloud resources operated by the providers supporting the WeNMR SLA and the resources provided by Open Science Grid (OSG). HADDOCK is an integrative platform for the modeling of biomolecular complexes. It supports a large variety of input data and it is also a core software in the BioExcel Center of Excellence to support COVID-19 researchers. Thanks to the cooperation already established between EGI, the Open Science Grid (OSG) and various high energy physics sites that committed to support COVID-19 related research, the HADDOCK platform was able to more than double its processing capacity, serving on average ~550 active users per months, 11,000 simulations related to COVID-19 (the equivalent of ~1.5 million HTC jobs, ~2.7 million CPU hours) on the EGI and OSC grid resources over the months of April to September 2020.

Read the full use case.

Giuseppe La Rocca is Community Support Lead at the EGI Foundation.

Gergely Sipos is the Head of Services, Solution and Support at the EGI Foundation.

Contributors: Davor Davidovic, Karolj Skala and dr. Drasko Tomic from the Rudjer Boskovic Institute in Zagreb, Alexandre Bonvin and his team in Utrecht (The Netherlands), Luca Fontanesi and his team from the University of Bologna (Italy), and Kallol Roy and his team from the Institute of Computer Science, University of Tartu, (Estonia).