At first glance, it might appear that HPC doesn’t have as important a place in helping the fight against the COVID-19 pandemic as the men and women who are on the front line in intensive care wards across the world. In the early days of any epidemic, a disproportionate amount of responsibility is heaped upon the dedicated people who run primary care (GP surgeries), work as emergency responders and those who work in any capacity in hospitals helping care for those who are worst affected by illness and disease.
But if you take a moment to look past the individuals who make up the medical teams, almost every new medicine they prescribe, every medical device they use and virtually every investigative test deployed has at some point had a significant amount of computational effort spent on it. This is where we see the everyday value from HPC, which we probably take for granted.
As the world continues to pull together to find an epidemiological solution to COVID-19, HPC will be at the forefront of the investigation and development of possible vaccines and drug treatments. HPC will be used on bioinformatic aspects of the disease and to develop better in silico methods for molecular science and the screening of patients and possible treatments. It’s clear that dealing with CORVID-19 will certainly be more of a marathon than a sprint and HPC will be playing its part.
The humanitarian and economic threat presented by COVID-19 is so great, that a worldwide coalition of traditional HPC centres and national labs, as well as the all the Cloud vendors are working to ensure that the worlds epidemiologists and the panoply of computational science disciplines that are now arrayed against the pandemic, have sufficient HPC resources to ensure that it won’t be lack of computational horsepower that defeats us.
HPC has so far been effectively deployed in the training of machine learning solutions to screening of patient CT scans, being able to identify the virus’s particular expression in lung lesions and even being able to distinguish between COVID-19 and other forms of pneumonia. HPC has also shown its capability in the running protein folding simulations which allow modern drug-docking (in silico screens) to identify candidate compounds which might disrupt COVID-19’s ‘spike’.
More recently a whole range of computational drug discovery workflows, including AI/ML driven ab initio models are being brought to bear on the problem. HPC has been successfully leveraged in the last ten or fifteen years to reduce the amount of costly and time-consuming in vitro studies required as part of the drug screening and discovery pipeline and I suspect that researchers and companies around the world will make great strides in utilising the power represented by both the larger capability systems as well as the amazing throughput and capacity represented by the Cloud vendors.
It is a truism that the great strides in technology are more often made in times of crisis and I think we will look back at the next couple of years and see that is the case for HPC. In my opinion, HPC hardware won’t have evolved significantly as a result of COVID-19 research efforts. But I do think there will be significant progress using in silico techniques alongside ML/AI systems in end to end bioinformatic and drug discovery pipelines, which will be enabled by using modern HPC systems and the burgeoning scale of the Cloud vendors.
“When you have a time of crisis what happens depends on what ideas are floating around, and what ideas have been developed, and thought through, and are made effective.” Milton Friedman
Dairsie Latimer – Technical Advisor