New Delhi:
An artificial intelligence (AI) model can predict which variants of the Covid-causing SARS-CoV-2 virus will likely bring about fresh waves of the infection, a new research has found. The model can detect around 73 per cent of the variants in each country that will cause at least 1,000 cases per 10 lakh people in the three months, following an observation period of one week, and over 80 per cent after two weeks, researchers said.
The team from Massachusetts Institute of Technology, US, and The Hebrew University-Hadassah Medical School, Israel, analysed 9 SARS-CoV-2 million genetic sequences across 30 countries, collected by the Global Initiative on Sharing Avian Influenza Data (GISAID). This was combined with data on vaccination rates, infection rates, and other factors.
The initiative “promotes the rapid sharing of data from priority pathogens including influenza, hCoV-19, respiratory syncytial virus (RSV), hMpxV as well as arboviruses including chikungunya, dengue and zika,” according to GISAID’s website.
The team used the patterns emerging from the analysis in building a risk assessment model based on machine-learning, an AI algorithm that can learn from past data and make predictions. Their study is published in the journal PNAS Nexus.
The researchers found that among the factors influencing a variant’s infectiousness, the strongest predictors were the early trajectory of the infections it caused, its spike mutations, and how different its mutations were from those of the most dominant variant during the observation period.
“These results support the hypothesis that the infectious new variants are those that acquire enough mutations which either can lead to reinfections or enable targeting new subgroups of the population that were naturally immune to previous variants,” the authors wrote in their study.
They said that the current models predicting the dynamics and trends of viral transmission do not predict variant-specific spread.
This study leverages variant-specific genetic data together with epidemiological information to provide improved early signals and predict the future spread of newly detected variants, the authors said in their study.
They said that the novel modelling approach could potentially be extended to other respiratory viruses such as Influenza, Avian Flu viruses, or other Corona viruses, and predict the future course of other infectious diseases as well.
Future research could explore how genetic and biological understanding of variant’s infectiousness and spread can be translated into predictive factors, that can be evaluated based on available data, the team said.
