Many pathogens of current relevance have coevolved with humans for thousands of years, while others have more recent origins in host-transfer events. In either case, there is often a cat-and-mouse game in which the virus seeks to avoid escape immunity and the host seeks to block transmission of the virus. Bedford combines viral sequencing, epidemiology and host-pathogen coevolutionary data to develop statistical models for predicting new flu antigen mutations that may allow the virus to overcome preexisting host immunity or even become a pandemic threat. These data are used to plot the genetic history of viral strains (a phylogenetic tree). A typical phylogenetic evolutionary tree has staggered branches. Influenza is an outlier to this trend. With each flu season, new mutant strains replace previous viral strains, leading to a ‘ladder-like’ phylogenetic tree (see figure).
Bedford is currently working on making seasonal predictions of influenza. Of all the different flu strains in the world, he’ll try to predict the one that will outcompete the others. This sort of research can be very useful for seasonal flu vaccination design and development. He analyzes sequences that are taken from around the world at different times and then designs a statistical model that describes those sequences and can explain the observed variation. An example would be to draw blood from a patient after they were infected with flu strain ‘A,’ and then test if their antibodies react to different strains of virus, e.g. flu strains ‘B’ and ‘C.’
“With these types of data,” Bedford explains, “we can get a pretty good idea of whether or not a person infected with strain A will then be protected against B, C, D, etc.”
In order to tackle these questions, certain information about the seasonal flu virus must be available; the phylogenetic lineages that emerge as the main circulating strains and mutations in HA and NA that promote immune evasion. Statisticians can then apply what they’ve learned about influenza evolutionary history to a model and then translate the model output to make predictions about future evolution.
“In order to deal with the shear amount and complexity of these phylogenetic data,” said Bedford, “we use a computational technique called Markov chain Monte Carlo.”
Markov chain Monte Carlo (MCMC) is a technique used to deal with the enormous number of phylogenetic trees that are possible. From this vast collection, MCMC effectively samples a smaller number of phylogenies that are supported by the sequence data. This technique allows integration of large sequence data sets with other experimental data. Bedford’s computational methods in conjunction with epidemiological, immunological and clinical studies on influenza genetics and pathology offer an innovative and unique opportunity to work toward the collective goal of reducing the global burden of influenza disease.
“The level of focus or activity in this field is very rare,” said Bedford while explaining why he came to VIDD. “I can’t think of a place that is similar, with so much going on in terms of making connections and collaborations. There are resources and data here that don’t exist elsewhere."
» Trevor Bedford faculty page
» Bedford lab website