Rapidly mutating viruses have posed one of the greatest threats to public health for decades. SARS-CoV-2 has claimed over six million lives within its first three years, while the influenza virus causes around 500,000 deaths annually. Against these pathogens, vaccines are one of humanity’s most effective tools. However, their VE is constantly challenged by the rapid evolution of viruses. New variants quickly emerge, adapt, and escape the immunities provided by prior vaccinations or infections.

The core issue lies in the antigen mismatch between the vaccine and the circulating viruses. As vaccines require months of development and manufacturing, by the time they are administered, they may no longer be effective against the now-dominant mutated strains. This phenomenon was evident with COVID-19 vaccines, where those used earlier viral strains demonstrated significantly lower VE against newer variants. Despite ongoing efforts to regularly update vaccine strains and renew vaccination programs, VE often remains suboptimal. According to the US CDC, the average VE of seasonal influenza vaccine in the United States in the 16 years from 2004-2020 was just 39.9%.

The time between beginning vaccine production and administration typically takes around 8-10 months, and vaccines should remain effective for at least 6 months post-administration. This means that to achieve high VE, vaccine antigens have to closely match the circulating viruses 12 months into the future. In other words, to design effective vaccines for tomorrow, we need accurate prediction of virus evolution today. However, conventional immunological approaches are limited by available samples, since one cannot test viruses that do not yet exist, and phylogenetic methods lack reliability and accuracy due to unreliable assumptions.

Predicting what mutations will occur is already difficult, but when and which mutations will become dominant is even more complex. To address this challenge, BethBio has developed new cutting-edge prediction models that overcome these limitations. Beth-1 is a robust prediction model that leverages bioinformatics, AI and computational biology to analyze publicly available viral genome data. This model could accurately forecast future evolution and identify the dominant circulating strains at specific future time points. By accounting for mutations that may occur during vaccine development and production, Beth-1 enables clients to design vaccine antigens that are better matched to circulating viruses, significantly improving VE and providing stronger immunity against emerging variants.

On the other hand, Beth-2 is more powerful for predicting mutations of newly emerged viruses such as SARS-CoV-2. By analyzing deep-sequencing genome data, the model analyzes the competitiveness of different mutations and predicts key mutations that are likely to be successful and dominant in future circulating strains. Beth-2 provides vaccine developers with sufficient lead time to prepare candidate vaccine antigens and build antigen libraries, enabling them to produce the next generation of boosters as soon as VE dropped significantly against new variants.

Beth-1 is the most extensively validated and best-performing prediction model for influenza virus evolution. In over 14 years of retrospective validation, Beth-1 used data up to year T to predict the dominant circulating virus in year T+1 and selected a representative available strain at year T. The results showed an average reduction of 7.7 amino-acid mismatches with actual circulating viruses for pH1N1 and 4.2 amino-acid mismatches for H3N2, compared to the seasonal influenza vaccine of that year. Beth-1’s prediction consistently performed better in every single season, corresponding to an estimated 4-40% increase in absolute VE.

As of 2024, BethBio has conducted 4 years of prospective predictions, where predictions were uploaded to blockchain before annual vaccine composition announcement, with affirmative results. Notably, Beth-1 can make predictions 24 months in advance and still has similar or fewer amino-acid mismatches with circulating viruses compared to traditional methods using predictions only 10–12 months ahead. A mice vaccine immunization experiment was also used to evaluate Beth-1’s prediction and showed superior neutralization titre in HAI assay compared to existing vaccines. Part of these results can be found in our publication on Nature Communications in 2024.

Beth-2 also delivers exceptional accuracy for predicting key mutations, with a median prediction lead-time of 8-9 months for SARS-CoV-2. The technology is currently capable of predicting 92.3% of key mutations in the RBD and 85.0% in the Spike protein prior to their dominance. It has over 85% AUC on RBD mutations and in retrospective analyses, could predict 28 of the 32 RBD mutations that defined the Omicron variant using data only up to the Delta variant – a feat many considered extremely challenging due to the likelihood of missing data.

With its state-of-the-art virus evolution prediction models, BethBio is at the forefront of enabling next-generation vaccines. Beyond influenza and SARS-CoV-2, BethBio is expanding its technologies to tackle other pathogens, including RSV and monkeypox, ensuring a better-prepared world for future health threats.

If you're a pharmaceutical company, vaccine producer, or research institute looking to revolutionize vaccine design and improve public health outcomes, partner with BethBio. Together, we can combat evolving pathogens and create novel effective vaccines that save more lives. Contact us at business@bethbio.com today to explore opportunities for collaboration.