HPC5 for EXSCALATE4CoV: supercomputer versus Coronavirus

Our supercomputer infrastructure HPC5 is the best-performing in use in industry anywhere in the world, with peak processing speeds of 51.7 petaflops per second. Combined with its predecessor HPC4, which is still running, the system can reach record speeds of 70 petaflops, meaning 70 million billion mathematical operations, per second. Its 3,400 computing nodes and 10,000 graphics cards allow HPC5 to process enormous quantities of data at once, while consuming very little electricity. The machine was designed and created for the energy industry, to reduce computing times and make it quicker to develop innovative ideas. One of its applications is in molecular modelling, the study of molecules using mathematical models that simulate their characteristics. Molecular modellers, who are hybrid scientists with expertise in chemistry, physics and mathematics, use complex algorithms developed by special programs to simulate molecules on the supercomputer, ridding them of the need to produce and study them physically in a laboratory. They can then study the behaviour and properties of well-known molecules and even predict how little-known ones will act. Applied to the field of energy, supercomputing and molecular modelling can select the best photoactive substances for Organic Photovoltaics (OPV) or Luminescent Solar Concentrators (LSC). In medicine and pharmaceuticals, they can be used to find new cures.

The spread of the SARS-CoV2 virus, which causes COVID-19, is putting the entire planet to the test. That’s why we need to combine our forces and all help look for solutions to this global struggle. Marrying the powers of HPC5 with the skills of the molecular modellers could provide real arms in the battle against this medical emergency. EXSCALATE4CoV, acronym of EXaSCale smArt pLatform Against paThogEns, seeks to find the safest and most promising drugs for fighting coronavirus. Steering the project is the Italian biopharmaceutical company Dompé, at the helm of 18 institutions and research centres in seven European countries, including Cineca in Bologna. In addition to Eni, other leading companies in their respective industries are also participating, specifically SAS and LEONARDO, which are involved in the artificial intelligence platform, cloud services and cybersecurity.

The first nucleus of the drug discovery platform was set up by Dompé Farmaceutici back in 2004. Since then, the platform has been continuously improved and integrated into the group's pharmaceutical research process up to the current 2020 version, which gave rise to the EXSCALATE platform, developed together with Cineca and Politecnico di Milano. The European consortium EXSCALATE4CoV was set up on this foundation and, with the support of the European Commission, led to the clinical trial of an anti-Covid molecule within a few months thanks to computing power and computer-aided design. EXSCALATE is also already equipped to handle Artificial Intelligence and IT developments such as quantum computing and neural microprocessors.

In October 2021, Dompé Farmaceutici's EXSCALATE platform received the SMAU Milano 2021 Innovation Award, for its capacity to respond to the need to accelerate the development of new solutions for patients by leveraging skills and technologies from other industries.

As part of the EU-funded [GC3] EXSCALATE4CoV project, in collaboration with Cineca we are offering HPC5’s computing power and our experience in molecular modelling. Our shared goal was to run simulations of the molecular dynamics in the proteins on the surface of SARS-CoV-2, which play a key part in the virus’s infection mechanism. This work lays the foundations for the next step, which is Dompé’s responsibility: screening 10,000 active ingredients in known pharmaceutical compounds, to find the most efficient for blocking the virus.

This was followed by the Fast Track phase, in mid-November 2020, during which the supercomputer was used in the molecular dynamic simulation of viral proteins considered relevant to the mechanism of infection of Covid-19. The screening was aimed at selecting the molecules with the highest anti-viral efficacy. Interactions between virus proteins and over 500 billion molecules have been reproduced on the computer using sophisticated molecular dynamics calculations to identify structures capable of attacking the virus with the aim of identifying those with the most effective pharmacological activity.

If this research bears fruit, we will have a clinically tested drug on our hands, completely ready for use. Molecular modelling uses very complex calculations that require enormous capacity. Without a powerful supercomputer to hand, it would take years to do all the necessary calculations. But HPC5 will let us study all the thirty or so proteins present in SARS-CoV-2 in just a few months. With HPC4, it would have taken two months to simulate the behaviour of a single protein. HPC5 can do the same in about 10 days. And happily, the race for new supercomputers shows no signs of stopping; after all, HPC4 was the most powerful industrial [GC4] computer in the world until last year, and HPC5 quadrupled its capacity. Our latest supercomputer also has the benefit of the ideal hardware to support Gromacs, an open-source programme that simulate molecular dynamics. This software divides the models into multiple parallel calculations, which is perfect for using all 7,280 of HPC5’s GPUs to their very best.

HPC5 was built using today’s energies to develop tomorrow’s energies. Its computing power has already proved its worth, letting us select and study new oil and gas fields faster and more accurately than ever, as well as develop innovative technologies for generating energy from renewable sources. One of these is ISWEC, which produces electricity from wave power. In fact, renewable energy is where the supercomputer revealed how useful it was for molecular modelling; the machine helped modellers pick the best photoactive compounds for luminescent solar concentrators (LSCs) and organic photovoltaics (OPV). Put to use in pharmaceutical research, HPC5 will also let us tackle new challenges and develop greater and greater skills in molecular modelling. Thanks to our part in the EXSCALATE4CoV project, we are gaining a ton of experience and building a valuable network of contacts in a sector that’s increasingly strategic for the energy industry.

In order to make the best possible use of the HPC5 supercomputer's extraordinary computing power, of course, equally exceptional skills are required. Eni's theoretical chemists, as experts in molecular dynamics, are involved in numerous projects, from the modelling of photoactive molecules for photovoltaic technologies (LSC and OPV), detergents and lubricants to improved oil recovery by means of organic polymers and aqueous emulsions. It is thanks to the synergistic work of computer scientists, organic chemists and Eni characterisers that such a wealth of experience could also be applied in the simulation framework of the EXSCALATE4CoV project. Within the company, the sharing of expertise has been the basis for many successes. Eni is a dynamic environment that puts its human capital at the heart of what it does and promotes both academic training and the circulation of company know-how.

The commitment on the part of Eni and its people to the fight against Coronavirus confirms the integration of the values expressed by the SDGs outlined in the United Nations Agenda 2030 in the company's mission and, consequently, in Eni's business model. Membership of the EXSCALATE4CoV project specifically reflects the company's commitment to Goal 3: Ensure healthy lives and promote well-being for all at all ages.

Molecular modelling has two great benefits to offer pharmaceutical research: it’s very fast at identifying new drugs and reduces the need for tests on patients. Studying and even predicting the pharmacological properties and side effects of active ingredients while reducing the times and difficulties of clinical trials is always important, but it’s vital in an emergency like the present one. So, now more than ever, computers may really take on a human face.