0
0
Scientists from The Universities of Manchester and Oxford have unveiled an innovative AI framework designed to detect and monitor new and concerning variants of COVID-19, offering a promising tool for future infectious disease surveillance.
Published this week in the journal PNAS, the groundbreaking study introduces a novel approach that combines dimension reduction techniques with a newly developed explainable clustering algorithm called CLASSIX, pioneered by mathematicians at The University of Manchester. This framework enables the rapid identification of groups of viral genomes that may pose a future risk, even amidst vast volumes of data.
Roberto Cahuantzi, a researcher at The University of Manchester and the paper’s lead author, emphasized the urgency of the endeavor: “Since the emergence of COVID-19, we have witnessed multiple waves of new variants, each presenting unique challenges. Our aim is to pinpoint these concerning variants early on, allowing for a proactive response, including targeted vaccine development.”
The study addresses the pressing need to enhance traditional methods of tracking viral evolution, such as phylogenetic analysis, which often requires extensive manual curation. COVID-19’s high mutation rate and rapid evolution necessitate efficient techniques for identifying potentially problematic strains.
The researchers leveraged the vast repository of genomic data available on the GISAID database, which contains nearly 16 million sequences of influenza viruses. By automating the analysis process, they were able to process 5.7 million high-coverage sequences in just one to two days using a standard modern laptop—a task that would be impractical with existing methods.
Thomas House, Professor of Mathematical Sciences at The University of Manchester, highlighted the significance of their approach: “The unprecedented volume of genetic data generated during the pandemic calls for improvements in our analytical methods. Our framework streamlines the analysis process, empowering researchers to identify concerning pathogen strains more efficiently.”
The AI framework works by deconstructing genetic sequences of the COVID-19 virus into smaller “words,” known as 3-mers, and grouping similar sequences based on their word patterns using machine learning techniques.
Stefan Güttel, Professor of Applied Mathematics at the University of Manchester, explained the novelty of their clustering algorithm: “CLASSIX is significantly less computationally demanding than traditional methods and provides textual and visual explanations of the computed clusters, enhancing interpretability.”
Cahuantzi emphasized the potential of their approach: “Our study serves as a proof of concept, showcasing the utility of machine learning methods in the early detection of emerging major variants. While phylogenetics remains essential, our methods offer scalability and efficiency, accommodating large datasets at a fraction of the computational cost.”
With the ongoing threat of emerging variants, this AI framework represents a significant step forward in our ability to monitor and respond to evolving infectious diseases, offering hope for improved public health interventions in the future.
