How AI and Machine Learning help to develop vaccines

Patrick Buckley    14 December, 2020

As Christmas approaches this year, we have all been gifted the great news that the Pfizer/BioNTech vaccine has shown to be both safe and effective in creating an immune response to COVID-19. Recently it has been approved for use both in the United States and the United Kingdom, with selected high risk British citizens becoming the first in the world to access the vaccine during early December.

In this post we briefly explain how Artificial Intelligence (AI) and Machine Learning technologies continue to play an increasingly important role in the development of vaccines.

How do vaccines actually work?

Vaccines create an immune response by exposing the patient to inactive, harmless virus particles known as proteins. Once the human body has been exposed to a virus, in an inactive form, it will develop antibodies. It is these antibodies which protect cells from becoming infected and, ultimately, prevent the patient from getting sick. Once these antibodies have been triggered once, the same immune response will be triggered every time the patient is exposed to the virus, allowing the patient to become immune.  

The role of AI.

When described in brief, the process of formulating a vaccine seems straight forward; simply identify the virus, extract inactive proteins that generate the immune response and you have a vaccine! Unfortunately, the reality is far more complicated. For an immune response to be activated, specific parts of the virus have to be exposed to antibodies. The challenge therefore is being able to identify these specific parts and understand their properties. Once these properties have been identified, scientists can extract the correct viral proteins that will trigger the best immune response. 

AI is becoming an increasingly useful tool in this process. As the COVID-19 pandemic started to grip the world back in January 2020, researches from the University of Stanford started to use Machine Learning solutions to identify proteins to include in a potential vaccine. Firstly, proteins of the SARS-CoV-2 virus were profiled, this is the virus which triggers COVID-19. Once the protein data had been collected, it was compared with data collected by researchers over many years on typical viral properties which trigger the antibodies to recognise common properties.

Once this data has been collected on a large scale, scientists are able to predict which viral proteins will trigger an immune response. This process would have taken far longer without the use of this technology and many of the insights gathered could not have been spotted by the human eye. This technology allowed  researchers to pass accurate insights and predictions to vaccine developers dynamically and quickly, allowing pharmaceutical companies to expedite the development of their vaccines without compromising on quality and safety.

This technology is currently limited by the lack of data to refer to. As AI & Machine Learning tools are increasingly used in vaccine development, more data will be collected, and scientists will have a deeper understanding of the viral protein properties which generate the best immune response.

Conclusion.

Vaccine development is an extremely complex and intricate process. Although the technology is still in its early days, Machine Learning tools have already contributed to the successful development of vaccines. As we continue to use Machine Learning in vaccine development, the availability and quality of the data on which it relies will improve. As the data becomes increasingly insightful, Machine Learning tools will become increasingly useful in vaccine development. 

To keep up to date with LUCA visit our website, subscribe to LUCA Data Speaks or follow us on TwitterLinkedIn or YouTube .

Leave a Reply

Your email address will not be published.