Bioinformatics Tackles the Flu

Illustration of influenza virus particles
Illustration: Ian Cuming/Getty Images

At first, it looks like a normal conference room: bland walls, long table, high-backed chairs. But then the lights dim, and a screen on the far wall begins to glow. I put on my 3D glasses.

A flu virus jumps off the screen. Not literally, thankfully. On the fourth floor of a non-descript building in Cambridge, Massachusetts, I sit with a team of scientists from Sanofi Pasteur, the vaccines division of the multinational pharmaceutical company Sanofi. In this room, the team uses virtual reality-like technology to analyze 3D images of the flu virus particles—currently a large purple and yellow orb dotted with bumpy red spokes—and the human antibodies that attack them. It’s a key part of Sanofi’s bioinformatics effort, which applies tools like structural modeling and genetic mining to improve influenza vaccine options.

“The 3D visualization lab allows us to perform visual analysis of these molecules at atomic-level detail,” says Eliud Oloo, manager of the Structure, Genomics and Informatics group at Sanofi Pasteur. He pulls up image after image of the virus onto the screen, sometimes zooming to the level of individual hydrogen bonds. That detail—and being able to see it in a realistic, 3D way—is helping to design vaccines that are more effective and last longer, says Oloo.   

The team’s goal is to help produce what they call a “broadly protective influenza vaccine,” or BPIV, which would protect an individual from many flu strains, even those that haven’t emerged yet, over multiple years. Sanofi Pasteur currently makes 21 approved vaccines—including a first-of-its-kind dengue vaccine approved in Mexico last year—and has another 12 new vaccines coming up through research and development.

The flu vaccine saves tens of thousands of lives, but there is plenty of room for improvement. Each year, the annual vaccine—produced not only by Sanofi Pasteur but many other pharmaceutical companies around the world—protects against three to four strains of the flu, but it’s not always right about which strains will be the most prevalent or the most deadly. Even when it is accurate, the vaccine only protects about 50-60% of people who receive it. And the World Health Organization estimates that easonal influenza still kills between 250,000 to 500,000 people worldwide each year. Outbreaks of especially infectious and deadly strains are always a threat.

Back in Sanofi’s visualization room, the set-up is simple yet effective. Two high-resolution projectors produce images for the left and right eye, to create the 3D effect. Because the team preferred a backlit setup, but still needed to fill a 6.5 square meter screen, they set up two long, horizontal mirrors to reflect light from the projectors over a long path toward the screen.

The team uses commercial 3D software programs such as PyMOL to visualize flu particles, and powers them through a local computer cluster. On-screen molecules are generated with data from a variety of public databases that contain DNA sequences of past and present strains of the virus. They also rely on internal data on viral and antibody structures produced by other Sanofi laboratories.

Oloo zooms in on a single flu particle to depict how a human antibody—a protein that binds the virus to neutralize it—latches onto a viral surface protein called hemagglutinin, the virus’s secret weapon that gloms onto a human cell so that the virus can infect it.

The team also uses the 3D system to observe how hemagglutinin mutates over space and time, as the virus evolves to avoid human antibodies. This can help the researchers identify pockets of the protein that change less frequently as the virus mutates—like aiming at a still target versus a moving target. So no matter how the virus mutates in the future, an antibody that binds that particular pocket can shut it down.

They also use the system to twirl and analyze different human antibodies, as some are better than others at binding and activating the human immune system. If the bioinformaticians can identify those antibodies, they can use them to narrow the list of vaccine candidates, potentially saving years in the laboratory and preclinical trials, says Oloo.

The team’s expectation is not to develop a “universal” vaccine, as other companies are pursuing, that could protect from all flu strains for a lifetime. Harry Kleanthous, head of North America discovery research for Sanofi Pasteur, doesn’t think that’s realistic in the near future. Instead, the company has doubled-down on what works now: vaccines that induce antibodies to neutralize the “business” end of the hemagglutinin proteins, so they can’t attach to and infect human cells.

“We think that you need the same kind of effective mechanisms that we rely upon today with current licensed vaccines, but we need to extend that coverage against many more circulating viruses,” says Kleanthous. “And we think it can be done.”

Those are grand goals, but with the help of 3D analysis, Kleanthous and Oloo are optimistic. The group has also used the visualization methods to study Zika, respiratory syncytial virus (RSV), and other infectious diseases.

Advertisement

The Human OS

IEEE Spectrum’s biomedical engineering blog, featuring the wearable sensors, big data analytics, and implanted devices that enable new ventures in personalized medicine.

 
Editor
Eliza Strickland
New York City
Contributor
Emily Waltz
Nashville
 
Contributor
Megan Scudellari
Boston
 

Newsletter Sign Up

Sign up for The Human OS newsletter and get biweekly news about how technology is making healthcare smarter.

Semiconductors

SIGGRAPH Asia: 3D models from photographs

Two presentations in a session on urban modeling here delved into generating three-dimensional models of buildings and streets from casual sets of photographs. Generating 3D models from 2D images isnâ''t a particularly advanced field, so these two new approaches definitely caught my eye. The state of the art requires a fair amount of user guidance to help the image-processing algorithms differentiate between a target object and visual clutter, such as trees, passing cars, and street signs. There's plenty of room for improvement in accuracy and detail, and users can always hope for a faster process and simpler interfaces. Currently, …

Advertisement