Currently, 25 vaccines to combat COVID-19 are in clinical evaluation, another 139 vaccines are in a pre-clinical stage, and many more are under investigation.
But many of these vaccines, even if successful, may not elicit the desired immune response in large sections of the population. This is because some people's antibodies will respond differently to vaccines that are supposed to stimulate virus-fighting T cells.
Calculating how much coverage a vaccine will have, that is, how many people it will affect to elicit an immune response, is a big part of the vaccine puzzle.
Bearing this in mind challenge, scientists at the Massachusetts Institute of Technology (MIT) on Monday presented a machine learning approach that can predict the likelihood that a particular vaccine project will affect a sufficient percentage of the population.
This does not mean that they can guarantee its effectiveness, but the work of scientists can help in knowing in advance whether a given vaccine has large gaps in who it can help.
The good news is that MIT researchers have used their approach to design a new COVID-19 vaccine on the computer that has much better coverage than many of the vaccines published in the literature this year. Now they are testing the project on animals.
The bad news is that there could be large gaps in coverage some of the existing vaccines have already been developed by companies and laboratories, according to one of the report's authors, David K. Gifford, who is in MIT's Department of Computer Science and Artificial Intelligence Laboratory.
"While they can protect more than 50% of the population, some people and the elderly may not be protected." said Gifford on ZDNet in an email when asked about the vaccines being tested.
It should be mentioned that the vaccines under development were not its immediate subject work. Most of these vaccines are closed projects, and no one knows exactly what they contain.
But Gifford and his colleagues designed vaccines from scratch, and then analyzed how effective they were, and worked with their findings on different vaccines whose composition is known.
Based on the above, one can conclude that there may be problems with other vaccines whose exact composition is not known.
We must keep in mind that any vaccine design from a computer, such as the one we mention here, is only the beginning of a process that can take years to pass on to animals and then to humans, to ensure that in addition to being safe (non-toxic), and effective, provide the community with a significant immune response.
However, the MIT project shows the computer model's ability to dramatically speed up the original project, looking for many possible combinations, a search that can take years as the world burns.
The research, entitled, “Computationally Optimized SARS-CoV-2 MHC Class I and II Vaccine Formulations Predicted to Target Human Haplotype Distributions,” was published in Cell Systems, by Cell Press. Authors include Gifford, Ge Liu, and Brandon Carter of the AI Lab, Duke University's Trenton Bricken, Siddhartha Jain, also of the AI Lab, and Mathias Viard and Mary Carrington, who have dual roles at Mass General and Frederick National Laboratory for Cancer Research in Maryland. (A post is also available on the MIT blog.)
