Weekly newsletter:
January 18-24 2021
Level difficulty:
Artificial intelligence in the service of vaccines
One of the first steps in the search for a vaccine consists of finding epitopes (or antigens, the region of the virus recognized by the immune system) that enable an efficient immune response to be triggered. The majority of studies dealing with SARS-CoV-2 have concentrated on the epitopes situated at the level of the spike (S) protein, the cell-surface protein of the virus most exposed to antibodies.
According to these studies, the immune response against the S protein is associated with low antibody levels and short memory immunity of B-lymphocytes (LB). Studies have shown that a T (LT) lymphocyte response following vaccination coupled with the activation of neutralizing antibodies gives more effective protection. Although the LTs cannot stop the virus from entering the cell, they confer protection by recognising viral antigens present on the surface of infected cells and thereby participate in the elimination of the virus (viral clearance). The presentation of antigens to the immune cells occurs thanks to the HLA system, proteins expressed on the surface of cells. The HLA protein population contains a certain diversity. It is important, therefore, to take account of this diversity when analysing the effectiveness of the viral epitopes that trigger an immune response.
The presentation of antigens to immune cells can be represented as follows:
To fight the pandemic, a vaccine must be able to protect a large majority of the world’s population. It is therefore important to select viral epitopes that allow a good LT response, by not focalizing uniquely on the surface proteins and in taking into account the diversity within the HLA system. This is what Danish and German researchers from the industrial group NEC attempted to do by using artificial intelligence to map out the epitopes within the totality of SARS-CoV-2 proteins.
These researchers had recourse to computer simulations of cell-surface antigen presentation in order to define the zones richest in viral epitopes that could be recognised by the numerous HLA systems covering the overall population. They were thus able to demonstrate that the most immunogenic regions are not uniquely situated in those zones exposed to antibodies, like the S protein. They then studied the potential of antigenic presentation of 3400 sequences of SARS-CoV-2. Some mutations in viral proteins seem to reduce their potential of presentation by the HLA system and this weaken their detection by the immune system. And on the other hand some mutations seem to increase their potential of presentation. The majority of the most immunogenic regions are preserved between the different variants, although some are eliminated and others appear. Finally, the scientists were able to select those epitopes that are not very susceptible to mutations and have few homologies with human proteins, in order to avoid autoimmune reactions.
This is what can be represented schematically as follows:
In conclusion, artificial intelligence can help the mapping of effective viral epitopes in vaccine research. However, it must be noted that these predictions are founded uniquely on computer simulations and need to be confirmed experimentally before they can be used in vaccine development.
