If you are new to the world of viruses, you might be wondering why you’d use a supercomputer to solve a medical problem. Is this all for show? Are IBM, Amazon, Google, Microsoft, and others just doing this as some sort of marketing ploy? Isn’t using the largest computers on Earth to investigate a tiny molecule a bit of an overkill?
The team, called, The COVID-19 High Performance Computing Consortium, as of this writing, consists of over 30 supercomputers. Members from private businesses, government agencies, and universities have donated the use of their supercomputers to any investigators seeking to solve the COVID-19 problem. Harnessing the power of this team of supercomputers, investigators are capable of doing 50 million calculations for each person on Earth in only one second.
Indirectly, of course, stopping the spread of the coronavirus benefits these participants because the longer it goes unchecked, the longer world economies will underperform and the more money these companies will lose. The virus hurts both private enterprises and governments. If the world wants to return to something close to normal, a widely available vaccine is a must; and the sooner this is developed, the better for everyone.
But this doesn’t answer the basic question: Why do you need a team of supercomputers? Take a look at this quote from a scientific paper about coronaviruses.
“Coronavirus replication entails ribosome frameshifting during genome translation, the synthesis of both genomic and multiple subgenomic RNA species, and the assembly of progeny virions by a pathway that is unique among enveloped RNA viruses.”
I know. That’s exactly what you thought all along. The point here is that coronavirus molecules are incredibly complex and their biological interactions with human cells are not fully understood. The first use of supercomputers may be to understand the protein structure of the virus and determine how it binds with and interacts with human cells. Here is a simplified diagram of a coronavirus molecule. Yes, this is the entity that has disrupted the life of every person on this planet.
It’s the molecule’s extreme complexity that poses such a problem for researchers. Drugs may already exist that could help battle this virus, but there is no way to test them all in clinical trials. There are also databases of millions of chemical compounds, some of which may go a long way towards mitigating the effects of COVID-19, but how can researchers know which ones may be effective?
The testing of the interaction of numerous pre-existing drugs with the virus is an unusually painstaking process. This so-called, ‘drug repositioning’ can be aided by simulating the interactions of these drugs with the virus through the use of sophisticated computer models. The use of supercomputers could dramatically speed up this testing process. Understanding gained in this way may not directly lead to a vaccine, but it could lead to the use of certain drugs as a sort of stopgap measure which could lower the severity of the disease in those patients who may be in high risk categories. The recently approved remdisivir is an example of such repositioning. Remdisvir was a failed Ebola virus treatment.
Supercomputers could also be used to identify those victims who have a genome that is specifically vulnerable to an attack by the coronavirus. If these people could be identified, they would know that they should take more precautions than those who do not have this vulnerable genome. They would, then, presumably, be more accepting of prolonged isolation. Such genomic research is normally time-consuming so supercomputer analysis would be a great help. Identifying genomes that are susceptible to the virus and those that seem to be resistant to the virus would allow researchers to select the appropriate people for a drug or vaccine trial.
There is a testing model that might be used to speed up the development of a COVID-19 vaccine. In this scenario, volunteers are given the vaccine and, subsequently, injected with the disease. If researchers had previously identified those genomes which were resistant to COVID-19, they would be the ones you’d want to select for such trials. You would want to screen your volunteers to make sure they are not among those who are most genetically susceptible to the disease; otherwise, you may end up killing your volunteers. Promoters of such an approach claim that such a “human challenge” trial could offer clear proof of a vaccine’s worth at blinding speed. We’re talking 2, 3 months.”
That may be a possibility, but, in most cases, it takes extensive trials and a lot of time to find out what, if any, side effects may develop from a potential treatment or vaccine. It would be much better if the side effects could be determined using computer models. This is, indeed, a possibility. In quite a few cases, a promising drug fails because the side effects are simply too severe. The risk of using the drug would then be too high except in extreme cases. If the models could predict side effects in advance, clinical trials could be implemented more quickly and with lower risk.
On the more practical, non-molecular side, better computer models could be developed using AI which would more accurately predict the spread of the virus. This would enable communities to prepare for outbreaks in advance or determine which policies to implement to protect their citizens.
There is, therefore, no doubt that the use of these supercomputers could go a long way towards finding a cure for COVID-19 more quickly. Once the complexity of the COVID-19 molecule is understood, scientists will have a better idea of how long it will take to develop a vaccine. Though it normally takes one to two years for such vaccines to be developed, tested, approved, and manufactured, it might be that the COVID-19 molecule is so complex that it could take years to unlock its mysteries. This is why all efforts need to be made to shorten the process. This is also why it is impossible to make firm predictions as to when a vaccine could be developed. Keep in mind that scientists have worked for decades on an HIV vaccine and have yet to develop one.
There is also the element of luck that can’t be easily dismissed. Oxford University researchers are currently testing a vaccine they developed. They are now at the point of testing it on human volunteers. By luck, they had already been working on coronavirus vaccines when COVID-19 showed up. They tweaked their research to focus on this particular virus and found encouraging results. If all goes well, they could have a vaccine ready by the fall of this year. That said, we know that other vaccines they developed did have side effects. These were not severe but they could have been. Oxford University does have its own supercomputer but their research has not been published yet so I have no idea what part computer modeling may have played in all of this,
To be honest, it would take quite a bit of luck to have a perfect vaccine ready by this fall. It could happen, but the odds are against it. Less than 10% of drugs that begin trials are ultimately approved. The following chart from the New York Times shows what they feel is the fastest that a vaccine could be developed.
So, even a one to two year timeframe is optimistic. That said, the use of supercomputers was not factored into this.
In short, no amount of ranting by politicians or news media will speed up the process. It all comes down to doing tried and tested scientific work. At this time, at least 95 vaccines are being worked on. The use of supercomputers will vastly speed up this process and, if I were to guess, a vaccine will be developed sooner rather than later. The only question is: How much sooner?