Follow Datanami:
December 1, 2020

New Algorithm Aims to Optimize Vaccine Distribution

In the last several weeks, Pfizer, Moderna and AstraZeneca have each announced that their vaccines are capable of achieving high efficacy. Pending approval, the first millions of vaccine doses are expected before the end of 2020 and billions are expected within 2021. Still, a couple of pressing questions remain: how many people will try to get vaccinated? And among those people, how will (initially limited) vaccine doses be distributed?

To that end, a team led by researchers from Washington State University and Pacific Northwest National Laboratory (PNNL) have developed a prioritization algorithm called “PREEMPT.” The algorithm is the latest fruit of a seven-year collaboration between the two institutions focused on scalable graph applications. 

When the pandemic hit, the team immediately pivoted to epidemiological applications. As a test, they built a tool that mapped out all the individual people in Portland, Oregon, as a series of interconnected nodes – a sort of synthetic social network. Using the algorithm to optimize vaccine distribution within this network, the researchers found that they could reduce total infections by up to 6.75 times and reduce peak infection by up to 98 percent.

“This formulation allows us to leverage principles from influence maximization – a well-known problem in network science,” said Marco Minutoli, a computer scientist at Pacific Northwest National Laboratory and lead author of the paper. “These types of strategies have not been studied for containing epidemics, especially in the context of uncertainties or lack of prior information. This application makes the work relevant to researchers, practitioners and decision makers in the epidemics domain.”

One limiting factor in the past has been sheer computational power – which now, of course, is less of an obstacle. The team ran their algorithm across 128 nodes on Summit, the 4,608-node Oak Ridge-based supercomputer that recently retained its second-place spot on the Top500 list of the world’s most powerful supercomputers.

“Speed is a critical factor, and we now have the ability to not just compute the largest number of possible solutions but also compute them quickly and accurately, so that critical problems are addressed in as close to real-time as possible,” said Mahantesh Halappanavar, a computer scientist at PNNL who also works at Washington State.

However, there remain a lot of obstacles before an algorithm like PREEMPT could be deployed in the real world: predominantly, that PREEMPT lacks many of the socioeconomic factors that have played major roles in the spread of COVID-19 to date.

“Social and economic determinants play a role in disease spread,” said Anil Vullikanti, a computer scientist at the University of Virginia Biocomplexity Institute who contributed to the research. “Algorithmic fairness is a big topic and something we all talk about. We want to make sure everyone in the community has the opportunity to get vaccinated without one community being left out.”

To read more about this research, read Washington State University’s Tina Hilding’s article here and read the research paper here.