null Puhti vs Coronavirus - Review of CSC COVID-19 Fast Track Outcome

Image: Jyrki Hokkanen and Mikael Kanerva

Supercomputer Puhti vs. the Coronavirus: Review of CSC COVID-19 Fast Track Outcome

Like many other organizations, CSC wanted to participate in the fight against COVID-19 and opened a fast track for projects researching this disease. Handling of applications for resources for such projects has been expedited and projects are granted a prioritized access to supercomputer Puhti. The fast track enabled CSC to speed up research when the need was greatest. The fast track also demonstrated the importance of permanent and high-quality research infrastructure. It is possible to respond quickly to acute needs when researchers have a functioning environment available to them.

A total of 15 research projects were selected for the COVID-19 fast track. These projects have investigated the airborne transmission of coronavirus, searched for ways to inhibit virus replication, studied potential drug molecules and their structure-guided optimization, screened for potential drug candidates, studied virus mutations and identified virus variants through sequencing. Almost all of the thousands of coronaviruses detected from Finnish patients have been classified on Puhti supercomputer.  

The COVID-19 fast track was opened in March 2020 and a third of the capacity of the Puhti supercomputer was initially allocated for the use of the fast track. Measured over the entire year, the use of the fast track ended up being much smaller.

The resource earmarked for COVID-19 research was nevertheless significant. In the early stages of the fast track, Puhti was the only supercomputer that CSC was using, and use of the computational resources of the fast track was highest before Mahti was taken into use. The opening of Mahti was originally scheduled for late April, but various technical difficulties delayed the launch by three months, and researchers did not get access to Mahti's computational power until late August 2020.

On the use of resources

A total of 15 research projects were selected for the COVID-19 fast track. CSC's usage statistics reflected the first signs of computing on the COVID-19 fast track in late March and the use continued moderately until mid-April. Intense computing began after 15 April and on a few days in April the fast track used up to 2/3 of Puhti's cores. 

Use of resources by the fast track was heavy until Midsummer and continued at a lower level through the summer. In early September there was one spike in usage but use mostly dwindled by the end of the year. One possible reason could be the introduction of Mahti on 26 August 2020, which multiplied the resources of CSC's high-performance computing resources and reduced the need for a special COVID-19 fast track.

Projects selected for the COVID-19 fast track used 26,367,453 billing units in 2020, which was 5.42% of the total usage of Puhti. 

The usage of Puhti supercomputer from March to December 2020. The usage of COVID-19 Fast Track is marked in red. Click image to enlarge.

Airborne transmission of the virus 

Janne Kuusela, Chief Physician at the South Savo Social and Health Care (Essote) who has been in the front lines in the fight against COVID-19 contacted CSC. He wanted research-based information about the transmission of the virus and how to protect against it. He was especially interested in the airborne transmission of the virus. 

CSC's Peter Råback recognized in this a task for computational fluid dynamics and contacted Aalto University Associate Professor Ville Vuorinen. Vuorinen had modeled turbulent flows and he has an excellent background in the analysis of the progression of droplets caused by coughing. Vuorinen was inspired and already the next day he set up a multidisciplinary group to study the problem.

CSC's resource allocation group headed by Juha Fagerholm quickly granted Ville Vuorinen's application two million billing units. On the third working day after Kuusela had contacted him, the work with COVID-19 began with the help of queue arrangements made by Ville Ahlgren and the first calculations were made with Puhti. 

– Hardly ever has an idea been transformed into a computational project of this size so quickly. The computational fluid dynamics community did not hesitate for a moment when demand emerged for their skills. CSC's role was to be a catalyst and enabler – researchers now have excellent facilities to do their work, Peter Råback says.

The group brought together by Ville Vuorinen studied how extremely small droplets that come from the respiratory tract when people cough, sneeze, or speak are carried by air currents, and the numbers of particles a person moving around spending time in public spaces might encounter. 

The modeling revealed that a cough from a person carrying the coronavirus raised the particle content in the nearby environment to such a high level that the risk of strong exposure was considerable as far away as four meters from the person coughing, and that the risk remained considerable for several minutes.

Dozens of researchers took part in the study, including experts in fluid dynamics, virology, medical technology, and infectious diseases. 

Researchers from Aalto University, VTT, and the Finnish Meteorological Institute conducted 3D modeling, each utilizing their own special skills and knowledge. The task of experts in virology and communicable diseases at the University of Helsinki is to interpret what the modeling means for coronavirus infections. The research group worked in interaction with doctors from Essote and THL experts in infectious diseases.

– The results of the study were published in the Safety Science in June 2020. This article was one of the first scientific publication to predict that the aerosol transport of the virus would be a very significant mechanism of transmission, contrary to what was previously thought. The result contributed to a radical change in the perception of the spread of the coronavirus, shifting the focus from surface infections to airborne infections. The results have had an impact on e.g. ventilation instructions and the use of masks, Ville Vuorinen says. 

That article is the most downloaded article in the journal and the study received a lot of media attention, both nationally and internationally. In the spring of 2020, there were about 1500 media news and the study was covered by the BBC and CNN, among others.

Visualizations have a great importance in scientific communications, especially when popularizing science. In this video Jyrki Hokkanen at CSC is visualizing how small droplets from the respiratory tract are spreading in the air.

CSC news: Finnish researchers join forces to investigate the airborne transmission of coronavirus 
CSC news: Modeling confirms: Isolating the ill and prioritising remote work are key strategies in combating the coronavirus 

Looking for medications

Ville Paavilainen at the University of Helsinki is using structural biology to study cotransins, which are small molecules that prevent entry of many SARS-Cov-2 proteins into the secretory pathway, which is an essential step in their biogenesis and could provide a mechanistically novel way to prevent virus transmission. Importantly, cotransins also inhibit biogenesis of e.g. SARS-Cov-2 Spike protein and its new variants, which may be important for controlling future outbreaks related to emergence of novel coronaviruses or other viruses.

Currently the Paavilainen laboratory and their collaborators are testing new versions of cotransin molecules, their effects against different SARS-Cov-2 strains and efficacy in animal models of the disease. 

@CSC online magazine: Structural biology guiding development of new SARS-Cov2 therapies 

The research group of Professor Ilpo Vattulainen at the Department of Physics of the University of Helsinki uses atomistic molecular dynamics simulation and machine learning to reveal the functional mechanism of mPro, the main protease of SARS-CoV-2. A protease is an enzyme which processes proteins and is responsible for the maturation of viruses. Preventing the main protease from functioning would also prevent the replication of the virus. Many medications have been suggested as ways of blocking the function of the main protease, but it is hard to advance the research if the functional mechanism of the protease is not first understood in detail. 

– An analysis of simulation results based on machine learning revealed that only a few key amino acids guide the catalytic activity of the mPro enzyme complex in a way in which the entire structure of the enzyme complex participates collectively in its activation. The study revealed several possible ways to prevent the action of the mPro enzyme, Ilpo Vattulainen says.

@CSC online magazine: Revealing the functional mechanism of the main protease of SARS-CoV-2

A group headed by Olli Pentikäinen, Professor of Medicinal Chemistry at the University of Turku, is studying the spike proteins on the surface of the virus which the virus uses to latch on to the ACE2 receptors of the cell to be infected. 

– Computer simulations made for the Sars-CoV-2 spike protein allowed us to identify several binding areas for small molecules affecting the attachment of several antibodies and the attachment of ACE2 receptors. The experimental testing of the molecules that potentially prevent interaction between the virus and the ACE2 receptor is now ongoing, says Elmeri Jokinen of Professor Pentikäinen's group.

In addition, researchers are automatizing a protocol that they have developed to speed up the evaluation of large-sized molecule databases in a virtual screening targeting the simulation structure of several target proteins.

Computer simulations in the study and screenings of the molecule databases aimed at finding candidates for medication were made on CSC's supercomputer. 

@CSC online magazine: Using computer simulations in search of drugs against SARS-CoV-2 

Professor Antti Poso and his research group (University of Eastern Finland) are using virtual screening for drug repurposing against COVID-19 infection. The screenings are typically based on three-dimensional molecular geometry. A large molecule library is scanned, and attempts are made to find molecules that might have the desired effect. This is done with the help of the docking of molecules and in some cases through pharmacophore searches.

The group of Research Director Markku Varjosalo of the University of Helsinki is experimentally measuring the proteins of the virus and of humans using a cross-linking mass spectrometer assay. Antti Poso and his group are amassing a database of existing pharmaceuticals, including those that have been clinically studied, but not taken into use. Funding for the study is coming from the Academy of Finland.

– We have made a platform to identify protein-protein inhibitors that prevent the protein of the virus from attaching to human proteins. We are screening clinically tested medications and the best candidates will be tested experimentally, Antti Poso says.

CSC’s computational resources enable this research. Varjosalo's group gives Poso's group sets of proteins on regular interval. The software that is used is Schrödinger Maestro, for which CSC has acquired an academic license. 

@CSC online magazine: Sharing enables screening 16 million structures in 7 minutes

Identifying virus variants through sequencing

Ravi Kant and Teemu Smura from the Emerging infections research group and the viral zoonoses research unit (University of Helsinki) are analyzing sequences of the SARS-CoV-2 virus with the help of the Puhti supercomputer and the in-built virtual analysis workflow.  For the data on viral sequences sequenced in Finland to be quickly available for visualization, the researchers set up the Auspice server in cPouta with whose help viral sequences can be openly shared. CSC's Kimmo Mattila assisted the research group in the installation of software, workflow, and the Auspice server.

University of Helsinki researchers and technology specialists have had a major role both in the research and in helping health officials in managing the pandemic, as more contagious variants have been identified through sequencing. 

Identifying virus variants requires significant computational power. There have been several hundred sequences to be analyzed every week and getting results quickly is important. Again, COVID-19 fast track has been helpful. Virtually all of the thousands of coronaviruses detected in Finnish patients have been classified with the Jovian and HaVoC pipeline in Puhti supercomputer, and new samples are constantly being analyzed.

The Auspice server
University of Helsinki press release: Sequencing adapted from research use to monitoring the coronavirus pandemic

Speeding up research when the need was greatest

The COVID-19 fast track will be closed before the end of May as its use tapered off. The launch of Mahti increased the computing capacity of CSC, and LUMI will bring a massive increase this year. Research related to COVID-19 can be supported in the future through regular resource allocation.

The fast track enabled CSC to speed up research when the need was greatest. The fast track also demonstrated the importance of permanent and high-quality research infrastructure. It is possible to respond quickly to acute needs when researchers have a functioning environment available to them, including sufficient computational capacity, data networks, databases, software, services, and support.

Read more:

Three blogs on COVID-19 were released on the CSC blog platform:

Peer-reviewed articles and manuscripts

Ville Vuorinen et al.: Modelling aerosol transport and virus exposure with numerical simulations in relation to SARS-CoV-2 transmission by inhalation indoors, Safety Science 2020

Ravi Kant et al.: Novel NGS pipeline for virus discovery from a wide spectrum of hosts and sample types, Virus Evolution 2020  

Phuoc Truong Nguyen, Ilya Plyusnin, Tarja Sironen, Olli Vapalahti, Ravi Kant, Teemu Smura: HaVoC, a bioinformatic pipeline for reference-based consensus assembly and lineage assignment for SARS-CoV-2 sequences. In peer review in BMC Bioinformatics.

E. M. Jokinen, K. Gopinath, S. T. Kurkinen and O. T. Pentikäinen: Detection of binding sites on SARS-CoV-2 Spike protein receptor-binding domain by molecular dynamics simulations in mixed solvents, in peer review.

K. Gopinath, E. M. Jokinen, S. T. Kurkinen, and O. T. Pentikäinen: Screening of Natural Products Targeting SARS-CoV-2-ACE2 Receptor Interface – A MixMD Based HTVS Pipeline, Front. Chem., vol. 8, p. 1084, 2020.

More about this topic » Go to insights and news »

Tommi Kutilainen