High-throughput Testing of SARS-CoV-2 Infection, Evolution and Immunity by Deep Sequencing
Protein engineering was used to profile the evolution of SARS-CoV-2 and its escape from the human immune system. Testing was carried out to ascertain how well the variants interacted with human cell receptors and their capability to evade neutralising antibodies. The data collected was used to train a deep learning model that is capable of predicting the efficacy of various therapeutic antibody candidates against different combinations of mutations.
The Need to Understand and Predict the Effects of Virus Mutations
When viruses such as the COVID-19 causing SARS-CoV-2 replicate, mutations occur. This leads to the emergence of new viral variants which differ in terms of their ability to bind to human receptors and infect human cells, their ability to evade the human immune response, the effectiveness of vaccines and antibody therapies and the rate at which the virus is transmitted throughout the population.
Upon SARS-CoV-2 infection, the immune system generates neutralizing antibodies targeting the receptor binding domain (RBD) of the virus. This hampers host cell invasion and, in turn, viral infection. Therefore, the RBD plays a crucial role in the ability of the virus to evade the human immune response. The immune system exerts a selection pressure on the virus such that the virus must continually evolve to evade antibody recognition and continue to spread. The need to understand and predict the effects of possible SARS-CoV-2 mutations on the ability of the virus to evade neutralising antibodies was and continues to be of paramount importance in staying at least one step ahead of the virus and ensuring the availability of effective antibody therapies and vaccines.
Leveraging Machine Learning to Outpace the Virus
This consortium used protein engineering to profile the evolution of SARS-CoV-2 and its escape from the human immune system. Rigorous testing was carried out to ascertain how well synthetic variants interacted with human cell receptors and their ability to evade neutralising antibodies. Leveraging the capabilities afforded by machine learning techniques, the data collected from these tests was used to train a deep learning model that is capable of predicting the efficacy of various therapeutic antibody candidates against different combinations of SARS-CoV-2 RBD mutations.
In addition, the trained model was used to explore the breadth of antibodies’ effectiveness against millions of potential viral genetic sequences. This enabled the team to uncover how combinations of different antibodies could be used to enhance resistance to viral evolution. These findings pave the way for next-generation antibody therapies.
The findings produced in this project have not only enhanced understanding of the evolutionary conflict that exists between viral and immune adaptation, but have also opened new avenues for the development of more effective treatments and vaccines.
Banner image above: Scientists working in Prof Reddy’s laboratory.
