BioNTech and InstaDeep have developed an AI early-warning system to detect dangerous coronavirus variants

BioNTech, the pharmaceutical company, and InstaDeep, a startup, have developed an AI early-warning system (EWS) that analyzes SARS-CoV-2 genome sequencing data and forecasts high-risk variants.

We, together with @instadeepai, have developed an Early Warning System as a new computational method detecting high risk #SARS-CoV-2 variants. In trial period, the system identified >90% of @WHO designated variants 2mths before their official designation. https://t.co/XJgH4LjHLE pic.twitter.com/vAQN2y9lVQ

— BioNTech SE (@BioNTech_Group) January 11, 2022

The new method combines structural modelling of the viral spike protein and artificial intelligence algorithms. This enables the system to quickly flag potentially dangerous coronavirus variants based on metrics that assess their fitness and ability to evade immune responses.

During trials, the EWS identified more than 90% of variants two months before they were designated by the World Health Organization.

The WHO-designated strains ‘alpha’, ‘beta’, ‘gamma’, ‘delta’, and ‘eta’ were detected by the system in the same week that scientists uploaded information about them into the database. The Omicron variant was flagged as high risk by the EWS on the same day its sequence became available.

Results underscore that the EWS can evaluate new variants in minutes and track their lineage in near real time. It can also scale as new data on coronavirus mutations emerges.

Co-founder and Chief Executive Karim Begir said that more than 10,000 new variant sequences are detected each week, and human experts cannot cope with the data at such a scale.

“For the first time, high-risk variants could be identified on the spot, potentially saving months of precious time,” he said.

According to BioNTech’s Chief Executive and founder Ugur Sahin, early detection of potentially dangerous coronavirus strains could become an effective tool for alerting researchers, vaccine developers, and health authorities, giving them more time to respond.

Earlier, in December 2021, a Canadian scientist announced the development of an algorithm that identified more than 1,000 potential COVID-19 treatments.

In March, British scientists developed an AI tool for diagnosing COVID-19 from cough sounds.

Subscribe to ForkLog News on Telegram: ForkLog AI — all the news from the world of AI!