Mutations lead to the evolution of a species, creating mutations that can adversely affect humans. The current Covid-19 pandemic is due to a mutation in SARS-CoV-2, a coronavirus, which jumped from bats to humans. Mutations may impede drug and vaccine production, as they may render antiviral drugs and vaccines ineffective.
It is impossible to predict mutations because there are too many unknown factors. If expressed in mathematical models, the relationship between mutation and quantitative causality is Mutation = f (x1, x2, x3 ...), where x1 refers to cause 1, x2 refers to cause 2, and so on. You can also assign zero to any cause, but the number of factors needed and the conditions required for mutation are unknown.
A University of Southern California School of Engineering research team has created a new Artificial Intelligence (AI) coronavirus treatment model that analyzes vaccine efficacy and helps accelerate the vaccine manufacturing process.
This method analyzes possible mutations in the virus and ensures that the most effective vaccines are rapidly identified. Studies have revealed that this machine learning (ML) model can shorten a vaccine design cycle from the typical months or years to just minutes or even seconds. The AI framework delivers vaccine candidates in seconds, allowing a rapid transition to clinical trials.
This model can also be adapted to help keep pace with the coronavirus as it mutates globally. When applied to SARS-CoV-2, the computer model swiftly eliminated 95 percent of the compounds and identified the most promising options.
The AI-centric method has identified 26 prospective coronavirus vaccines. Out of the 26, scientists selected the 11 best vaccines to fashion a multi-epitope vaccine that attacks the coronavirus spike proteins that the virus uses to attach and enter host cells.
Vaccines target the epitopes (areas of infection) and disrupt the spike protein, counteracting virus replication. Using machine learning, engineers can formulate a new multi-epitope vaccine for a brand new virus in under a minute, and check its quality in under an hour. This technique is crucial during this pandemic stage as the coronavirus mutates in populations spread across the world.
The accompanying schematic diagram shows a traditional in silico vaccine design process. Researchers use various in silico tools to predict the HTL, B-cell, and CTL epitopes on the whole virus proteins, and the best regions for virus protein extraction occurs after an exhaustive examination of its physicochemical properties. This process is necessary to create an effective vaccine. The outcome carries a substantial overhead as a single in silico vaccine design tool can only extract a single prediction goal. For example, numerous researchers employ the popular BepiPred25 B-cell epitope prediction tool for B-cell epitopes prediction. BepiPred is limited to single-step prediction. Multiple predictions are beyond the scope of present-day tools.