Research into Corona: no time to lose

Researchers who needed computing facilities for research into covid-19 could apply for these last spring via an accelerated application procedure at SURF and NWO. For this article, we asked some of them about their research, how they used SURF's services and what their (preliminary) results were.

Molecuul SARS-CoV-2 door iSO FORM LLC

“Covid-19 and pneumonia are two very different diseases on the molecular level”

Felipe Vieira Braga

Felipe Vieira Braga, immunologist Amsterdam UMC

“Since the first cases of covid-19, it has been initially diagnosed as a type of pneumonia, especially due to its respiratory symptoms. Now there is plenty of clinical data showing that covid-19 is actually a much more complex disease, that affects different organs. However, it is largely unknown how the disease really compares  to pneumonia on the molecular level. Our research  tried to answer that question: how similar is the immune response against covid-19 compared to other classical pneumonias?

My research is very technologically guided. I am an immunologist who specializes in single cell RNA sequencing analysis, which is a very data intense genomic analytical method. One of the wonders of these technologies is it can be easily applied to different systems to address a variety of research questions. So it was the obvious choice to address our question. We wanted to analyse some genomic datasets, which required more computational power than what we had available. 

I therefore made use of SURF's HPC Cloud service. That provided me with an interactive analytical platform, where I could run my analyses and easily visualize them, to better understand the biology behind my data. The immune response in covid-19 proved to be very different from the immune response in classical pneumonia. These are two very different diseases, on the molecular level. We will follow up to better dissect these differences.”

Artificial intelligence detects covid-19 in CT scans

Erdi Calli

Erdi Calli, PhD student Radboudumc

With a so-called CO-RADS score, a radiologist assesses a CT scan on the basis of a set of diagnostic criteria to predict the probability of a covid-19 infection. “We have automated the CO-RADS system using an artificial intelligence pipeline. We collected CT scans of patients from three institutions in Amsterdam, Bremen, and Nijmegen. We created a "Release Bot" that collected the CT scans from the hospital archives, and via Research Drive created a daily release to share it with all parties involved in the development of this artificial intelligence pipeline.

This had to be done automatically and very quickly because around 50 people were working on different parts of the pipeline, such as data collection, annotation, preprocessing, and predictive models. Also, since it was the peak of the pandemic, we had to be able to find, process, and move a lot of data immediately!

Since the project required the utmost urgency, we applied for the covid-19 fast track. We were looking for a fast, safe and affordable option for storing and sharing high volumes of data. Also, we had requirements such as: the system should let us communicate via programmable interfaces, command line tools or libraries. And we wanted to have an option to add users with various permissions. One option was to use Amazon S3 for this purpose, however it was very expensive for the volume of data that we were working with. We found out that Research Drive meets all these requirements.

In the end, we have created systems that periodically query the data from the hospital archives, as well as algorithms that can make a diagnosis using this data. We have published multiple papers on detecting covid-19 infection using CT scans and our software is available on”

CT-scans van longen met negatieve CORADS-uitslag

Negative CO-RADS results

CT-scans van longen met positieve CO-RADS-uitslag

Positive CO-RADS results

Contamination can happen quickly in a cycling peloton

Bert Blocken

Bert Blocken, full professor Civil Engineering TU Eindhoven and KU Leuven (Belgium)

"Our research focused on the appropriate 'social distance' when two or more people move together in the same direction, at the same speed. Such as runners in a group or cyclists in a peloton. We used the national supercomputer to create computer simulations and animations.

The now familiar 1.5 metre distance applies to two people standing still, without wind, and facing each other. For example, when they are talking to each other. This distance is a good compromise between avoiding the exchange of large drops and what is practically feasible. But that distance no longer applies when people move together in the same direction at considerable speed. Then you have to keep a much greater distance. Because if you cycle in the slipstream of someone who sneezes or coughs or just exhales, you can breathe in those drops and aerosols. Our research shows how droplets are exchanged in a group of runners or cyclists. And that contamination in a peloton can happen very quickly as soon as one rider is infected.

This is common sense in itself, but apparently common sense is rare during an international pandemic. Runners and cyclists were accused of spreading corona and I received a storm of criticism. But then other scientists in aerodynamics confirmed our results. Today, international top virologists, epidemiologists and microbiologists agree with us. Olympic national organisations have consulted us for advice in relation to the Tokyo Olympics in 2021".

You can read more about this study in a blog by Bert Blocken on (penultimate story).

Wielrenner die aerosolen verspreidt

The importance of open science during a pandemic

Alejandro Lopez Rincon

Alejandro Lopez Rincon and Aletta Kraneveld, researcher and full professor at the Institute of Pharmaceutical Sciences, Utrecht University

Lopez Rincon: “We analysed the genome of the corona virus, which is very time-consuming and costly on our regular computer.  Therefore, we applied for computing time on the Cartesius supercomputer and the Lisa Compute Cluster. With these systems, we only needed a few hours to do what would take weeks on a regular system. 

We focused on two research projects: we wanted to see if we would be able to better detect the virus using deep learning and machine learning. Secondly, we wanted to investigate if we would be able to make a differentiation between symptomatic and non-symptomatic patients.”

Kraneveld: “Alejandro, who is a is a computational scientist, found databases that enabled him to take a closer look at the genome of the virus. His unique approach was that he used the whole genome sequence, whereas normally you only sequence a small part. He found another piece of RNA which is much more specific for the current virus going around. The results of the other project suggest that the absence of symptoms may be attributed to a mutation of the virus. It’s a great example of open science and sharing data: if the data hadn’t been available, we wouldn’t have made these discoveries, which are so important in the current pandemic situation.”

Article: Classification and Specific Primer Design for Accurate Detection of SARS-CoV-2 Using Deep Learning

Author: Josje Spinhoven

(Header image: CoV-2 molecule, credit: iSO-FORM LLC)

This article was published in SURF Magazine 2020-03