As the coronavirus continues to spread across the globe and reports of more contagious variants emerge from the U.K., South Africa, and elsewhere, it is clear that vaccine design and development could benefit from a speed boost.

To help accelerate the vaccine design cycle, researchers from the University Southern California (USC) have developed DeepVacPred, a novel AI-based in silico multiepitope vaccine design framework that incorporates a mixture of in silico immunoinformatics and DNN technologies.

"This AI framework, applied to the specifics of this virus, can provide vaccine candidates within seconds and move them to clinical trials quickly to achieve preventive medical therapies without compromising safety," said Paul Bogdan, associate professor of electrical and computer engineering at the USC Viterbi School of Engineering and corresponding author of the study, in a statement. "Moreover, this can be adapted to help us stay ahead of the coronavirus as it mutates around the world."

To compile a dataset for training the DNN, the researchers used data from the Immune Epitope Database (IEDB), a public database containing detailed information on immune epitopes related to infectious and immune-mediated diseases, and the Virus Pathogen Resource, a complementary repository of information about pathogenic viruses. Data on the SARS-CoV-2 genome and spike protein sequence were taken from the National Center for Biotechnical Information.

In addition to predicting potential vaccine subunits from the available SARS-CoV-2 spike protein sequence, DeepVacPred was also trained to analyze different components of the immune response including linear B-cell epitopes, cytotoxic T lymphocyte (CTL) epitopes, and helper T lymphocyte (HTL) epitopes in the subunit candidates. The results were then used to identify the most promising components for constructing a multiepitope vaccine against the SARS-CoV-2 virus.

DeepVacPred was able to eliminate 95% of the candidate compounds, leaving only the most promising candidates. In less than a second, it predicted 26 potential vaccine subunits that could work against SARS-CoV-2, a process that would have taken days using traditional methods. The researchers then narrowed down these 26 vaccine subunits to a set of 11 to be used to construct a multiepitope vaccine targeting the virus's spike proteins.

The result was a multiepitope vaccine candidate against SARS-CoV-2 containing an adjuvant, 11 subunits with 16 B-cell epitopes, 82 CTL epitopes, and 89 HTL epitopes.

The researchers then used a suite of bioinformatics tools to test the vaccine candidate's toxicity, physicochemical properties, and other characteristics. For example, the IEDB population coverage analysis tool was used to analyze the vaccine candidate's performance across human populations and verify that the vaccine subunits are able to cover a wide range of human populations.

Finally, the vaccine candidate was tested against three common RNA mutations in the spike protein or in other related proteins. For this experiment, the researchers entered the mutated protein sequences as input into DeepVacPred. In all three instances, the vaccine subunits predicted by DeepVacPred were the same as the original virus, demonstrating that the RNA mutations had no influence on the final multiepitope vaccine.

"The proposed vaccine design framework can tackle the three most frequently observed mutations and be extended to deal with other potentially unknown mutations," Bogdan said.

Confronted with a new virus, DeepVacPred would be able to construct a new multiepitope vaccine for it in less than a minute and validate its quality within an hour, according to the authors.

Do you have a unique perspective on your research related to vaccines or infectious diseases? Contact the editor today to learn more.