Beyond aesthetics: a visualization tool for better vaccines

From the Bedford lab, Vaccine and Infectious Disease Division

Haute couture and the flu have more in common than you might think.

Like fashion trends, the influenza virus is capable of rapid change. Like hem lengths, the dominant version of the flu virus changes every year. That’s why, when fashion week rolls around in the fall, we need to get an updated flu shot that has been specifically formulated to protect us from the versions of the virus circulating that year. In other words, last season’s flu vaccine just won’t cut it.

Just as your sweater in a shade of cerulean represents countless hours and jobs in the fashion industry, the flu shot that goes into your arm is a product of cutting-edge knowledge about the circulating variants generated by scientists. Among other things, two key pieces of information determine the vaccine composition for a particular year: the genetic composition and relatedness of circulating viruses (i.e., how related the circulating variants are to each other) and serological data indicating antibody efficacy.

In a new paper published in Frontiers in Bioinformatics, researchers from Trevor Bedford’s lab in the Vaccine and Infectious Disease Division at Fred Hutchinson Cancer Center developed a tool allowing users to simultaneously view serological and phylogenetic data for multiple strains of flu viruses.

Nextstrain, an analytical and visualization toolkit previously developed by the Bedford lab and collaborators, already allowed researchers to visualize phylogenetic data for different influenza variants. With Nextstrain, users could select a single vaccine candidate and compare its antigenic distance (a value indicating antigenic closeness among virus variants) to other circulating viruses. However, comparing multiple vaccine candidates required users to view the phylogenetic data for each candidate separately, alongside a static heatmap indicating antigenic distances between multiple vaccine candidates.

Researchers in the Bedford lab, including Jover Lee and James Hadfield, the lead authors of the current study, realized that a tool allowing users to simultaneously visualize phylogenetic and experimental serological data would facilitate the choice of effective vaccine candidates. “We developed an interactive tool [within Nextstrain] that allows researchers to visualize data from their wet lab experiments alongside the genealogy of their experimental organisms,” said Dr. John Huddleston, a staff scientist in the Bedford lab and the senior author on the study.

A flowchart of tools used to view data for influenza vaccines. Top left: binders with stacks of papers. Top right: interactive phylogeny and static heatmap. Bottom: phylogeny and graph showing serological data.
Historically, serological data indicating influenza-vaccine efficacy was presented in tables. In the 2010s, scientists developed tools that allows decision makers to visualize antigenic evolution in a phylogenetic context. Now, the new visualization tool described in this paper allows users to simultaneously view phylogenetic and serological data for multiple strains of influenza variants. Image provided by Jover Lee

To develop a streamlined and user-friendly interface, the authors of the paper turned to visual-design principles. For example, previous software had demarcated antigenic distance with color. However, design principles suggest that quantitative data, such as antigenic distance, is better indicated via positionality (for example, along the x- or y-axes on a graph). On the other hand, nominal data, such as the names of phylogenetic clades, are effectively coded with color.

The result is a streamlined panel that shows antigenic distance on the x-axis, with clades on the phylogeny indicated with color. In addition to being more user friendly, the new visualization tool reveals the underlying distributions of the raw data, an important information that the static heatmap obscured.

The authors tested the new visualization tool with two case studies, using data that informed the composition of the fall 2009 flu vaccine. They found that the new tool revealed biologically relevant patterns in the data that had been masked in traditional summary statistics.

“We show that this kind of visualization can inform decisions about the composition of the seasonal influenza vaccine and reveal patterns in the raw experimental data (such as multiple peaks in a distribution) that have been invisible to previous methods of displaying the same information,” said Dr. Huddleston.

The results from the case studies made the authors question what other patterns in experimental data have been hidden by approaches that rely on summary statistics, and whether automated detection of these patterns could be linked to data visualization methods like the ones described in this paper.

Dr. Huddleston said the team remains motivated by this important question: “How can we empower researchers to explore their organism’s genealogical relationships through the lens of the experimental data instead of the other way around?”

“To answer these questions, we will need to collaborate closely with other experimentalists on the visualization of their data and encourage retrospective analyses of previously published work,” Dr. Huddleston said. In the future, the researchers plan to expand the functionality of the visualization tool to allow users to explore the genealogy of their organism by filtering and selecting relevant experimental data.

In the meantime, users can try out the new measurements panel by heading over to the Nextstrain website, which provides a demo of the tool and an example workflow.

This work was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health.

Lee J, Hadfield J, Black A, Sibley TR, Neher RA, Bedford T, Huddleston J. 2023. Joint visualization of seasonal influenza serology and phylogeny to inform vaccine composition. Frontiers in Bioinformatics. 3:1069487.