BioNTech Develops Successful SARS-CoV-2 Early Warning System with InstaDeep

Global virus and disease spread, coronavirus
[Source: ismagilov/Getty Images]

BioNTech and U.K.-based artificial intelligence (AI) company InstaDeep have developed an accurate early warning system to enable detection of high-risk SARS-CoV-2 variants at the earliest possible opportunity to help health systems and government authorities prepare suitable control measures.

The system works by analyzing SARS-CoV-2 sequence data and applying AI calculated immune escape and fitness metrics and over the last year has successfully managed to identify more than 90% of World Health Organization designated variants before they were officially announced.

“Despite a relatively slow mutation rate in the human coronaviruses, since the emergence of the human coronavirus SARS-CoV-2 in Wuhan in December 2019, over 250,000 different missense variants (as of November 25, 2021) have been identified in the protein-coding viral sequences deposited in the GISAID (Global Initiative on Sharing All Influenza Data) database and associated with multiple lineages,” write the combined research team in a recent preprint describing the work.

“While most mutations either reduce the overall fitness of the virus, or bear no consequences to its features, some individual or combinations of mutations lead to high-risk variants, with modified immune evasion capabilities, and/or improved transmissibility”

The early-warning system to detect these high-risk variants was trialed for 11 months between September 2020 and November 2021, carrying out a weekly data scan during this period. The model was trained with data from variants sequenced up to the beginning of the previous month.

Two types of data feed into the model. First, structural modeling data that assesses how the viral spike protein will bind with the host cell receptor, a score on how likely the variant is to escape the immune response is then generated. Second, the AI system creates predictive models based on sequence data of the variants collected around the world and deposited in the GISAID database.

The system was able to identify 12 out of 13 high risk variants from Alpha to Omicron listed by the World Health Organization an average of two months (58 days) before they received their official designation. Often, this was when only a few sequences (less than 0.5% of the weekly average) had been recorded of the variant in question. For context, the majority of WHO announcements and designations occurred when around 18% of the weekly average of new variant sequences was due to that variant.

The only variant not identified in advance, Delta, was underrepresented in the GISAID database feeding the model.

BioNTech, now famous for developing the first COVID-19 vaccine in collaboration with Pfizer, began a partnership with InstaDeep in late 2020. The two companies have formed an AI Innovation lab to develop various AI and immunology related projects including improved drug design, advanced data analytics (such as this SARS-CoV-2 monitoring project) and manufacturing and supply chain optimization.

Going forward, the early warning tool should help give health authorities more time to plan and respond to the emergence of such variants.

“Early flagging of potential high-risk variants could be an effective tool to alert researchers, vaccine developers, health authorities and policy makers, thereby providing more time to respond to new variants of concern,” said Ugur Sahin, CEO and co-founder of BioNTech, in a press statement.

“More than 10,000 novel variant sequences are currently discovered every week and human experts simply cannot cope with complex data at this scale. We’ve addressed this challenge by deploying the powerful AI capabilities of InstaDeep’s DeepChain platform combined with BioNTech’s SARS-CoV-2 know-how and technology,” added Karim Beguir, Co-Founder and CEO of InstaDeep. “We are happy to make our research work publicly available and, most importantly, look forward to its continued real-world impact.”

This site uses Akismet to reduce spam. Learn how your comment data is processed.