Frankfurt The Mainz company Biontech and its British partner Instadeep want to significantly improve and accelerate the prediction of dangerous virus variants with an AI-based early warning system. The system that the partners presented on Tuesday afternoon is a calculation method based on sequence data that is available worldwide. According to Biontech, the system is able to identify high-risk variants of the Sars-CoV-2 virus two months earlier than conventional methods.

“With the new calculation methods that we have developed in recent months, we can analyze the sequence information of the spike protein and classify new variants according to their predicted immune escape and ACE2 binding score,” explained Biontech boss Ugur Sahin. Both factors determine how strongly and how quickly the pathogen can spread. The ACE2 binding is relevant to the ability of viruses to penetrate body cells.

The early warning system combines structural modeling of spike proteins with artificial intelligence (AI) in order to
Detect and monitor high-risk variants of SARS-CoV-2; on average, over 90% of the
variants declared by the WHO are identified two months before official designation
The study introduces a new method that combines publicly available sequence information with predictive
Analysis methods are combined in order to effectively identify and monitor high-risk variants and thus better.

The early warning system is fully scalable for new virus variant data
The study is available on the preprint server BioRxiv and was published in a scientific journal with
Peer review system submitted.

MAINZ, Germany and LONDON, United Kingdom, January 11, 2022 – BioNTech SE (Nasdaq: BNTX, “BioNTech”) and InstaDeep Ltd.

(“InstaDeep”) today announced the development of a new calculation method that analyzes and analyzes sequencing data available worldwide.

Predicts high-risk variants of SARS-CoV-2. The early warning system (Early Warning
System, “EWS”) is based on metrics calculated by artificial intelligence (AI) on immune evasion and viral fitness.
The new method combines structural modeling of the viral spike protein with AI algorithms to identify potential high-risk variants within recognized as such by less than a day. To do this, the system uses information entered in SARS-CoV-2 sequence databaseswill. The calculation is based on metrics for evaluating viral fitness (e.g. ACE2 and spike protein interaction of the  Virus variant) as well as the properties for immune evasion. The companies validated these predictions using specially generated experimental data as well as publicly available data.

During the trial period, the system had more than 90% of the virus variants identified by the World Health Organization (WHO) (Variants of Concern, “VOC”; Variants of Interest, “VOI”; Variants Under Monitoring, “VUM”) detected an average of two months in advance. The from Variants identified by the WHO alpha, beta, gamma, theta, eta and omicron were recognized by the EWS in the same week that their. Sequence was first uploaded. The omicron variant was released the same day its sequence became available as. Classified high risk variant. The hydroforming variant observed in France was also evaluated by the EWS. The results show that the the ability of this variant to bypass the immune system is similar to the Omicron variant, but in that it has a significantly lower level of fitness.
This makes them less of a concern given the current data.

The results of the study underline that the EWS is able to recognize new variants within minutes and their risk is almost in monitor in real time. It is also fully scalable for new virus variant data.

“With the new calculation methods that we have developed over the past few months, we can obtain the sequence information of the spike protein analyze and classify new variants according to their predicted immune escape and ACE2 binding score, ”said Prof. Ugur Sahin, CEOand co-founder of BioNTech. “The early detection of potential high-risk variants could be an effective tool to help researchers, Alert vaccine developers, health authorities and policymakers in a timely manner, allowing more time to initiate appropriate to have measures against questionable new virus variants. ”

“Currently, more than 10,000 new sequence variants are discovered every week. Experts cannot handle this complex data on this scale cope with more. We met this challenge by leveraging the powerful AI capabilities of the DeepChain platform from InstaDeep combined with the SARS-CoV-2 know-how and the technology from BioNTech. For the first time, high risk variants can appear on recognized at the point, which can save months of valuable time.