BioNTech and InstaDeep develop system to detect potential high-risk SARS-CoV-2 variants
BioNTech and InstaDeep announced the development of a new computational method that analyzes sequencing data available worldwide and predicts high-risk SARS-CoV-2 variants.
The early warning system (EWS), developed in collaboration with BioNTech and InstaDeep, is based on immune escape and fitness scores calculated by artificial intelligence.
The new method combines structural modeling of the viral spike protein with artificial intelligence algorithms to quickly flag potential high-risk variants entered into SARS-CoV-2 sequence repositories in less than a day, based on scores assessing their fitness as well as their immune escape properties.
The companies confirmed these predictions using experimental data generated in-house and publicly available data.
During the trial period, the system identified more than 90% of variants designated by the World Health Organization (WHO) on average two months in advance:
- variants of concern (VOC)
- variants of interest (VOI)
- variants under monitoring (VUM)
The WHO-designated Alpha, Beta, Gamma, Theta, Eta and Omicron variants were detected by the EWS in the same week their sequence was first uploaded.
The Omicron variant was identified as a high-risk variant on the same day its sequence became available. The IHU variant, observed in France, was also assessed by the EWS: it showed immune escape properties relatively similar to Omicron, but with significantly lower fitness, which makes it less dangerous given current data.
The study results highlight that the EWS can assess new variants within minutes and track variant lineages in near real time. It also scales fully as new variant data emerges.
“Thanks to the advanced computational methods we have been developing over the past months, we can analyze Spike protein sequence information and rank new variants according to their predicted immune escape and ACE2 binding score. Early identification of potential high-risk variants can become an effective tool for alerting researchers, vaccine developers, health authorities and policymakers, thereby providing more time to respond to new variants of concern”
The early warning system is based on two approaches: structural modeling of the interaction between the viral spike protein receptor-binding domain (RBD) and the host cell receptor, and assessment of the variant's effect on immune evasion, as well as AI-based predictive modeling to extract information from hundreds of thousands of virus variants recorded in global sequence repositories.
The EWS calculates an immune escape score and a preliminary fitness (transmission potential) score. While the immune escape score alone was already a highly predictive risk factor, combining these two scores into a Pareto score provided the best estimate of the risk associated with a given virus variant.
The higher the score, the greater the risk that the strain will affect global health. The EWS approach ranks SARS-CoV-2 variants by their ability to escape immune response and their fitness potential based solely on existing data, and therefore does not depend on a 'wait and see' approach.
“More than 10,000 new variant sequences are currently detected each week, and human experts simply cannot keep up with complex data at this scale. We solved this challenge by deploying the powerful artificial intelligence capabilities of InstaDeep's DeepChain platform in combination with BioNTech's SARS-CoV-2 expertise. For the first time, high-risk variants could be detected on the spot, potentially saving months of precious time. We are pleased to make our research publicly available and, most importantly, hope for its further real-world impact”
The EWS was able to distinguish WHO-designated variants from those without a designation over an 11-month period, highlighting the viability of the computational model for determining variant origins.
The weekly analysis conducted between September 16, 2020 and November 23, 2021 identified 12 of 13 WHO-designated variants, with an average lead time of 58 days before the variants received their designation.
For variants from Alpha to Mu, on average only about 25 cases had been recorded at the time they were flagged by the EWS. This contrasts with WHO announcements, which came when more than 1,500 cases had been recorded. The EWS identified Omicron, on the day its sequence was first uploaded, as the variant with the highest immune escape level among more than 70,000 variants detected between early October 2021 and late November 2021, and also assigned it a high fitness score.
Biopharmaceutical New Technologies is a next-generation immunotherapy company developing new treatments for cancer and other serious diseases. The company uses a broad range of computational discovery and therapeutic platforms to rapidly develop new biopharmaceutical products.
Its broad portfolio of oncology product candidates includes individualized and off-the-shelf mRNA-based therapies, innovative T cells with chimeric antigen receptors, bispecific checkpoint immunomodulators, targeted cancer antibodies and small molecules.
Building on its deep expertise in mRNA vaccine development and its in-house manufacturing capabilities, BioNTech and its collaborators are developing several mRNA vaccine candidates for a range of infectious diseases alongside a diverse array of oncology drugs.
Founded in 2014, InstaDeep is today the leading EMEA-region provider of enterprise decision-making AI products, headquartered in London with offices in Paris, Tunis, Lagos, Dubai and Cape Town.
Drawing on its extensive know-how in GPU-accelerated computing, deep learning and reinforcement learning, InstaDeep has created products such as the new DeepChainTM protein design platform, which solve the most complex challenges across various industries.
InstaDeep also collaborates with global leaders in the AI ecosystem, such as Google DeepMind, Nvidia and Intel. The company is part of the Intel AI Builders program and has received elite partner status from Nvidia. InstaDeep was recently named by CB Insights as one of the world's 100 most promising AI startups for the second year in a row.
