Jump to content
Powered by

How can influenza virus prediction be improved?

Every February, the World Health Organisation (WHO) publishes recommendations on the composition of influenza virus vaccines for use in the upcoming influenza season. The WHO’s decision is based on observations and laboratory tests as well as experience and intuition. The particular type of influenza virus that is likely to be circulating in a given season has previously been mainly a matter of speculation. Richard Neher from the Max Planck Institute for Developmental Biology in Tübingen, in cooperation with scientists from the UK and the USA, has developed software that takes a less speculative glimpse into the future.

Richard Neher, an independent research group leader at the Max Planck Institute for Developmental Biology, uses his knowledge as a physicist to predict the evolution of influenza viruses. © Helmine Braitmaier

"It takes a good six months to produce an influenza vaccine. It is therefore important to anticipate well in advance the influenza virus strains that are likely to circulate in the upcoming winter season," says Dr. Richard Neher adding that "seasonal influenza viruses can mutate so rapidly that the strains that cause the well-known influenza epidemics differ from year to year. This is why we need new influenza vaccines every year." Sometimes, the WHO picks the wrong virus, as was the case in February 2014. Shortly after the WHO had published its recommendations and vaccine production for the upcoming winter season (2014/2015) was well underway, a new influenza virus subtype appeared and became the dominant virus that season. The influenza vaccine that had been developed was therefore only effective in about 23 percent of people immunised. The effectiveness of good influenza vaccines normally ranges from 50 to 60 percent.

Neher and his colleagues have come up with a method to predict the evolution of influenza virus strains from information contained in genealogical trees. Such family trees are constructed by comparing the sequences of the viruses' haemagglutinin surface receptors and showing how each virus strain is related to others. The greater the similarity between haemagglutinin gene sequences in two virus strains, the more closely the strains are related. "We are especially interested in family tree sections that give rise to evolution of a large number new virus subtypes," says the 35-year-old researcher.

Which influenza subtype is the fittest?

Immune pressure on influenza viruses forces the viruses to adopt strategies that enable them to elude increasing immunity among humans. New virus strains with altered surface molecules can escape human immune responses as virus neutralising antibodies have not yet been produced in the population at large. This enables these subtypes to cause seasonal epidemic outbreaks. Random sequence errors that occur as the genetic material is copied during the propagation of the viruses leads to further subtypes that will be more or less successful. "Our mathematical method analyses the branching patterns of the reconstructed genealogical trees in order to infer the relative fitness of the internal nodes," explains Neher. Fitness relates to the ability of a certain strain to survive. Neher's method therefore spots expanding clades which are likely to dominate the future population.

"Our approach does not require historical data," says Neher. Neher and his colleagues simply screen worldwide influenza databases for the genetic fingerprints of all influenza virus variants that are circulating at a specific time somewhere on the planet and use these sequences to reconstruct genealogical trees. This information is then fed into the prediction software. Neher and his colleagues demonstrated the performance of the software using historical data from 1995 to 2013. They were able to predict successful progenitor strains for seven of the years; they failed for three of the years, and the predictions were of intermediate accuracy in the remaining years. Although intermediate accuracy was not the best possible result, it was still better than if the strains had been selected completely at random.

Predicting the evolution of viruses

Neher’s algorithm calculates from the genealogical tree of influenza viruses the virus subtype that is most likely to circulate in the upcoming winter season. © nextflu.org

Over the past few months, Neher and a colleague from Seattle have developed an interactive tool that allows users to colour the tree according to biological fitness, mutations or geographical origin of the influenza virus variants. "When a new influenza virus variant comes from the Far East, which is the epicentre of influenza epidemics in Europe, we have to spend more time looking at this variant than at a variant that originates in South America," says Neher. In order to track the spread and evolution of influenza viruses, the scientists have also included influenza virus data from the past twelve years in the new software. Every week, Neher's team adds to the prediction software influenza virus sequences that have been newly deposited in public databases.

However, Neher is not just seeking to use his prediction software for influenza viruses; he also wants to use it to predict the evolution of other viruses, as well as of bacteria and cancer cells that undergo rapid change under the selective pressure of the human immune system. This works too, as Neher's algorithm only requires the information stored in the genealogical trees for the predictions; information about the biological characteristics of the viruses or cells is not required. Neher is currently studying the evolution of the AIDS-causing immunodeficiency virus (HIV) in individual patients. "I would like to find out which viruses can evade the human immune system and which are destroyed by the human defence system," says Neher.

Neher, who is a physicist by training, is able to look at microscopic and macroscopic phenomena such as the effect of the behaviour of single viruses on the entire population using statistical methods. He also writes a blog (neherlab.wordpress.com) where he provides regular updates on his research.

Website address: https://www.gesundheitsindustrie-bw.de/en/article/news/how-can-influenza-virus-prediction-be-improved