The URV predicts future variants of Covid with an algorithm

A group of researchers from the Department of Biochemistry and Biotechnology of the URV have designed an artificial intelligence model that anticipates the next mutations of SARS-CoV-2 and predicts what the future variants of the coronavirus that in 2020 generated the emergency will be like sanitary.

"This is a methodology that learns from the hundreds of thousands of virus genomes that have been obtained in the last two and a half years and that are in the Gisaid database," says researcher Gerard Pujadas, professor of Biochemistry and Molecular biology.

"The methodology also serves to prepare us for other pandemics," says Santi Garcia-Vallvé, researcher and professor of Biochemistry at the URV

The study will be published in a few days in the International Journal of Molecular Sciences. The report has been led by Professor Santi Garcia-Vallvé, from the Chemoinformatics and Nutrition Research Group. It is part of the thesis of doctoral student Bryan Saldivar. "Predicting the next mutations is very useful to be able to design antiviral drugs, or even vaccines that may be useful for future variants of SARS-CoV-2," explains Saldivar.

There are changes that happen by chance but others are recurring and follow patterns

Garcia-Vallvé details the origin of the research: «There are a series of mutations due to chance or errors when the virus replicates and that are very difficult to predict. But we and the rest of the scientific community saw that there was a type of mutation that occurred more than other changes and that it could be caused by us, by enzymes that we need and that can be a defense mechanism against infection."

The scientists saw that there were recurring mutations: they occurred more than once and independently. From there came the development of the predictive model. “We reserved four important places in the genome, such as the part that encodes the Spike protein – the gateway to infection – and we saw if we could predict mutations in these areas”, points out Garcia-Vallvé.

The model has been tested with more than 4.6 million SARS-CoV-2 genomes

The group found that the method had predictive power. "The model is robust enough to predict long-term mutations, as some false positives within a limited time frame become true over a long period of time," the study explains. Somehow, the work was anticipating. «More and more genomes are accumulating in the databases. We used the data up to January 2021. Some of the mutations that had to be recurrent, according to our patterns, we did not observe at the time, but we did see that they occurred later," Garcia-Vallvé points out.

Anticipation

Anticipating future strains can be key to science, since sometimes the mutation causes an optimization of the virus. Names like Delta and Ómicron have become commonplace during the different waves of the pandemic.

"Predicting mutations is useful to be able to design antivirals or even vaccines," explains Bryan Saldivar, a doctoral candidate at the Faculty of Chemistry of the URV

In addition, it is a gateway to locate more pharmacological components with which to combat not only Covid. “This methodology helps us to be more prepared, also for future pandemics. Other coronaviruses work the same way, so it can be useful to predict which mutations could be observed," says Garcia-Vallvé. "These results can be used to find antiviral drugs," says the publication, the result of dealing with a huge volume of information. To create the model, 877,000 virus genomes were analyzed and 4.6 million to test it.

Leave a Reply

Your email address will not be published. Required fields are marked *

Go up