COVID-19 vaccine development

In December 2019, SARS-CoV-2, the virus that causes COVID-19, was completely unknown. Indeed, it wasn’t even clear whether it was a coronavirus or another variant of the influenza virus. Unfortunately, here in the UK, the authorities assumed the latter, but that is another story.

As infection numbers started to rise exponentially in many parts of the world, the race was on to identify this particular pathogen and find effective treatments and ways of controlling it.

The priority was to find drugs that could be used to reduce the severity of the symptoms. Typically, it takes between ten and twenty years to bring a new drug to market. However, it was very soon clear that the world could not wait that long. Even a new vaccine was likely to take several years to develop and many more to test to ensure it was both safe and effective.

Before developing a vaccine could occur, it was necessary to identify what type of beast this was. This initial work identified the virus as a coronavirus – basically a sphere with spikes. This meant it was more closely related to the common cold than to influenza.

Vaccines that can be used against influenza have been around for many years, so they are well-tried and tested. But, unfortunately, no effective vaccine existed to be used against colds. One reason is that their morbidity rate is relatively low. Another is that there are around 200 known variants.

Researchers in both pharmaceutical companies and universities worldwide started to look for methods that could be developed to combat this disease which was clearly starting to get out of hand.

It is not clear whether the major laboratories decided to adopt different approaches independently or via formal or informal discussions. Whatever the mechanism, the result is that there are several vaccines available that work in different ways. For some of the vaccines produced, it is unlikely that they would be available today had the developers not used AI/ML techniques to whittle down the potential options to a manageable level before deciding which to take through to the next stage of development.

This was only the first stage, as AI also proved invaluable during the clinical trials. To test the efficacy of the vaccines, it was necessary to select candidates for testing in areas where the pandemic was at its worst to obtain a sufficiently large sample size to be reasonably confident that the vaccine was effective and, at the same time, safe to use. Identifying these locations required the services of another AI tool.

Dimitris Bertsimas, MIT, brought his group of 25-plus doctoral and master’s students together to discuss how they could use their skills in ML and optimisation to create new tools to aid the world in combating the spread of the disease. Their models were soon generating accurate real-time insight into the pandemic.

The model had to be capable of capturing behaviours related to measures put into place during the pandemic, such as lockdowns, mask-wearing, and social distancing and the impact those measures had on infection rates. By July 2020, it could make predictions on 120 countries and all 50 US states on a daily basis.

More at: bit.ly/BehindC19