The COVID-19 pandemic has highlighted the need for rapidly developing effective countermeasures targeting new and emerging pathogens, the varied organisms that cause disease. If the single virus that causes COVID-19 has been able to wreak this much havoc on us all, clearly we’re in this battle for the long term. And we know that safe and effective vaccines are the first line of defense, as a critical component in a robust response to any current, emerging or future biological threat.
The problem is that vaccine design, testing and manufacturing are monumentally time consuming and expensive. The remarkable advances of COVID-19′s mRNA vaccines were only possible because initial research into the concept of mRNA vaccines had begun decades earlier, in the 1980s, and at long last the science was ready to be put to work. We cannot count on that stroke of luck for the next pathogenic invader.
A pharmaceutical company’s cost to develop a single vaccine can be up to $68 million, with failure rates as high as 94%. It’s a high-risk, high-reward operation. Only the toughest players succeed. They choose which vaccines to develop based on which are most likely to work, along with having the largest potential markets. Thus, vaccine development typically starts with multiple candidates following a lengthy linear workflow to mitigate these costs and risks.
But along the lines of working smarter, not harder, our team at Los Alamos National Laboratory, with Defense Threat Reduction Agency support, is developing a machine-learning tool to predict the most suitable vaccine technologies for a given pathogen. This will increase the rate of success and reduce the number of initial vaccine candidates required. That’s millions of potentially wasted dollars put instead toward more likely candidate projects.
Through a new, multi-laboratory consortium led by Los Alamos, we are developing the Rapid Assessment of Platform Technologies to Expedite Response (RAPTER) tool to systematically test seven different vaccine platforms with a variety of viral and bacterial pathogens to measure what works and what doesn’t. RAPTER will use artificial intelligence to look for patterns in the way these vaccines work so that when the next threat emerges, we know which one to use. We won’t have to throw everything at the wall just to see what sticks.
The key is that each vaccine platform generates a specific immune response based on its design features. Similarly, the host immune system needs to generate a distinct immune response to survive an infection by a pathogen.
When the next new pathogen emerges, we are developing RAPTER to predict which vaccine platform will best match the immune response required to protect someone from that pathogen. Pathogen lock, meet immunological key.
This has been a tough approach in the past, as the complicated nature of these immune responses and high variability in existing data has limited our ability to directly compare vaccine platform technologies. But advances in machine learning allow us to draw conclusions from a variety of data sources and identify an immunological profile for each vaccine platform and for entire classes of pathogens.
Targeted experiments with uniform, standardized protocols will be performed to fill the inevitable data gaps, ensuring a direct comparison of seven vaccine-platform technologies while also validating the accuracy of RAPTER’s predictions. We anticipate that the RAPTER tool developed, tested and validated in this research effort will revolutionize the future of vaccine development by rapidly down-selecting a suitable vaccine platform for any viral and bacterial pathogen.
Jessica Kubicek-Sutherland, Ph.D,. is a molecular biologist at Los Alamos National Laboratory.