13 April 2006

Can Computing Power Thwart Avian Flu?

By Rusty Rockets

Avian flu isn't such a hot news story at the moment, but experts say the threat of a global pandemic still remains. That's why researchers and health authorities are still toiling away trying to produce effective vaccines and feasible contingency plans. One such team of scientists from the Los Alamos National Laboratory in New Mexico, the University of Washington and the Fred Hutchinson Cancer Research Center claim they have simulated on a supercomputer how avian flu is most likely to spread. This information, says the team, will provide politicians, authorities and emergency workers with the best possible chance at containing the virus while an effective vaccine can be produced.

Deadly influenza pandemics are nothing new. In 1889 the Asiatic Flu was first reported in Bukhara, Russia, and within a year it covered the globe. The Spanish flu, arguably the most infamous of the influenza pandemics, occurred in 1918 and was the product of the H1N1 type virus. Nobody is exactly sure how many fell victim to the deadly virus, but estimates range from 25 to 50 million people worldwide. Asian Flu was the next pandemic to hit, over the period 1957 to 1958. Caused by the H2N2 type virus, it was first detected in China, but gradually spread to include the United States where approximately 70,000 deaths were reported. The Hong Kong Flu, first reported in (duh!) Hong Kong in 1968, was initiated by H3N2 and caused around 34,000 deaths in the United States. This particular influenza A virus still does the rounds today, with outbreaks varying in strength.

Is there any rhyme or reason to these outbreaks, and are we due for another influenza pandemic caused by the H5N1 type virus any time soon? While nobody is entirely sure, the World Health Organization (WHO) and many other scientists are strongly in favor of a better-safe-than-sorry approach. Medical historian Dr. Jim Leavesley, author of Mere Mortals, has his own views on the current H5N1 debate. "I think that these epidemics take time to develop, by that I mean a year or two, so we're not out of the woods, so to speak. And when it does it will be worldwide, with very little, if any, resistances."

If the history of influenza pandemics tells us anything at all, it tells us that once a flu of this caliber breaks out, there is little chance of it not claiming a significant number of lives. As Dr Leavesley says, there will be little resistances to the virus should it become human-to-human transmissible. Humans don't have a natural defense against viruses, because we only build up anti-bodies when we get the flu. "For instance," says Leavesley, "from the 1918 flu that killed 50,000,000 people, virtually all the people that had survived it have died off. So, they may well have had antibodies; and they don't pass those antibodies on, of course. It's a self-limiting thing, because after a few years the virus begins to build strength again, because there's no one growing up with it."

But despite the inevitability of pandemics, the history of our relationship with the influenza virus also shows something else. For whatever reason, when we look back at the figures of the last pandemics, there are noticeably less deaths, despite the virus' global spread. It may just be that the different strains of the virus are becoming less deadly, but it could also mean that we are getting better at responding to such catastrophes. Effective localization and containment not only keeps mortality rates down, but also buys scientists some time to identify the strain and prepare a vaccine. But in the highly mobile world of airplane travel, would we be able to contain a virus effectively, or are we now on the brink of yet another pandemic that will rival the 1918 Spanish Flu?

Using a highly sophisticated simulation model, scientists at Los Alamos have made some predictions about how avian flu may spread in today's environment of worldwide connectivity. The team explains how they used a large-scale, stochastic (a non-deterministic random function generator) simulation model to predict what possible path a human-to-human transmissible version of the H5N1 virus might take. The simulation is run on the Los Alamos supercomputer known as Pink. For boffins interested in the guts of the supercomputer, it has a 1,024-node (2,048 processor) LinuxBIOS/BProc "Science Appliance" running Clustermatic 3, which is the largest single-system image Linux cluster in the world. Pink's nodes have dual 2.4 GHz Intel Xeon processors (Pentium 4) with 2 gigabytes of memory per node (so it probably doesn't fit on a desktop).

The simulation model run through Pink's systems relates to how the virus would spread among a population of 281 million people over a period of 180 days, which is visualized using a city-and-census-tract-level picture. The stats used in the simulation match United States census demographics on such things as how workers get to work, in addition to randomly assigning simulated citizens to households, workplaces and schools. "We are only computing the probability of any person becoming infected on any given day, and a 'roll of the dice' is needed to decide whether they are infected or not," said researcher Timothy Germann. Further random projections filtered into the simulation include a portion of the population infected with the virus who do not develop any symptoms (but are infectious), as well as variations in regard to incubation periods.

Using Department of Transportation travel data allows the team to factor in how the pandemic progresses in regard to airlines and other forms of long-distance travel. This particular aspect of the simulation does not factor in alternate mitigation strategies, but does underestimate the amount of long-range travel (by 10 percent) once travel advisories are issued. But even in this scenario, the pandemic spreads and peaks (100 or more cases per 1,000 persons) rapidly after 90 days. "In the highly mobile US population, travel restrictions alone will not be enough to stop the spread; a mixture of many mitigation strategies is more likely to be effective than a few strictly enforced ones," said Kai Kadau, of Los Alamos' Theoretical Division.

The researchers say that the data gleaned and analyzed from such a simulation - including the impact of interventions, anti-viral therapy, school closures and travel restrictions - allows a more holistic approach to a virulent outbreak of influenza. The team hope that predicting the path of the virus will lead to improved levels of containment of the virus, which will in turn lead to more time to develop a vaccine. "Based on our results, combinations of mitigation strategies such as stockpiling vaccines or antiviral agents, along with social distancing measures could be particularly effective in slowing pandemic flu spread in the US," said co-author Ira Longini.

The team's results from the simulation also show that using even a moderately effective vaccine will ensure a better outcome than waiting for a match-specific vaccine that may take longer to develop. "Based on the present work, we believe that a large stockpile of avian influenza-based vaccine containing potential pandemic influenza antigens, coupled with the capacity to rapidly make a better-matched vaccine based on human strains, would be the best strategy to mitigate pandemic influenza," say the researchers.

The team is bullish about supercomputers as an effective way to predict and control the spread of an epidemic. "Computer models serve as virtual laboratories where researchers can study how infectious diseases might spread and what intervention strategies may lessen the impact of a real outbreak," said Jeremy M. Berg, director of the National Institute of General Medical Sciences. "This new work exemplifies the power of such models and could aid policymakers and health officials as they plan for a possible future pandemic," he added.