During the COVID-19 pandemic, the world lived in dread of new viral variants emerging which could reinfect individuals, spread more rapidly, and hinder the effectiveness of vaccines and other preventative measures. With EVEscape, however, novel strains of concern can be predicted and planned for so that they will no longer be able to take us by surprise.
The problem with escape mutations
Immunity to a virus is achieved principally by making antibodies: Y-shaped proteins produced as a result of either a previous infection or a vaccine. These molecules work by recognising and binding to specific molecules, called antigens, on the surface of viruses. The antibodies then either neutralise the virus directly or mark it as a target for destruction by other immune cells.
However, viruses, especially RNA viruses such as SARS-CoV-2, have a set of characteristics which allow them to evolve extremely rapidly. The mutation rate of their genetic code is very high, and they replicate at a fast rate and in huge quantities. The result of this is that any advantageous mutation which arises can spread very quickly and without difficulty.
These are often ‘escape mutations’ which result in a change in the viral antigen so that it is no longer recognised by the antibodies. As a result, the virus is able to ‘escape’ the immune response, allowing renewed infection and illness in individuals who have previously been infected with the same virus. In addition, current vaccines may no longer be effective against these new strains and may need to be updated, or completely redesigned in response, both of which may be long and costly.
What is EVEscape?
To tackle this problem, researchers at the University of Oxford and Harvard Medical School have developed a new AI tool, EVEscape, to predict which new variants are likely to emerge, before the mutations have even arisen. This is possible because escape mutations have to fulfil certain criteria in order to create a new variant. Specifically, they always change the part of the virus which antibodies recognise (the antigen) in a way that does not reduce the ability of the virus to survive and reproduce.
EVEscape was trained on pre-pandemic coronavirus genetic data and information about the 3D structure of their spike proteins, the primary antigen that is targeted by antibodies. Astonishingly, the AI tool returned a list of possible escape mutations, including every known mutation that led to a new variant of SARS-CoV-2 during the pandemic.
Predicting the future
This means that EVEscape has the capacity to inform vaccine design and other interventions before the spread of new viruses, helping to prevent future outbreaks. This represents a huge advance compared to the methods already available: attempts at similar computational frameworks have thus far been far less accurate.
For example, ‘deep mutational scanning’, the prediction method used throughout the pandemic, while accurate, requires antibodies from infected patients and laboratory experimentation. This means it has a limited ability to make long-term predictions and facilitate early interventions.
While of course it’s too late to use EVEscape for the COVID-19 pandemic, it has the potential to reduce the threat of emerging viruses which have been identified as possible sources of global pandemics, such as Lassa and Nipah.
AI at the frontiers of biosciences
A perfect storm of advances in computing have converted AI into one of the hottest topics in many fields of research, with the biosciences being no exception. Thanks to its ability to make inferences based on huge quantities of data that are too complex for manual analysis by humans, its potential applications are widespread, and it will be exciting to see where it is utilised next.
