Log in
Pitt Logo

Font Size


News Archives

Supercomputers Alter Science

A recent NYTimes article explores the impact supercomputing is having on science. Says author John Markoff, "Computing is reshaping scientific research...[it] has made it easier to share research findings, and that in turn has led to an explosion of collaborative efforts. It has also accelerated the range of cross-disciplinary projects as it has become easier to repurpose and combine software-based techniques ranging from analytical tools to utilities for exporting and importing data." Markoff further states what MIDAS research teams have long known: "Computer power not only aids research, it defines the nature of that research: what can be studied, what new questions can be asked, and answered."

For MIDAS, simulating large-scale epidemics requires the computational power of supercomputers, provided in large part by the Pittsburgh Supercomputing Center, a joint venture between the University of Pittsburgh and Carnegie Mellon University. Increasingly complex problems require complex models, and a popular tool for MIDAS researchers is what is known as Agent-Based Modeling (ABM), where individual persons are represented as autonomous "agents" who move within a given population. Agent populations are constructed using real data (for example, the U.S. Census) and individuals may be assigned to households, their children to schools, and may go to workplaces with a specific commuting distance, and other demographic factors.

Because Agent Based Models (ABM) represent an entire population inside the computer,they require large amounts of memory. For a simulation of H1N1 spread in the DC metropolitan area, MIDAS’s ABM included 7.4 million people, requiring seven gigabytes of memory. “This is a shared-memory problem,” notes Shawn Brown, head of MIDAS's Computational Team, referring to massively parallel systems that allow each processor to access all the memory without message passing. Brown and the MIDAS team are now working on scaling up their ABM model to cover the entire United States, incorporating a population of 300-million agents and requiring from 74 to 300 gigabytes of memory. 

Read the full story from the NYTimes here, and visit the Pittsburgh Supercomputing Center's homepage here.


MIDAS-Affiliated Student Wins Award

sunjoo low res 2009Michelle Sunjoo Lee, a member of MIDAS Investigator Bruce Lee's team and a junior at North Allegheny High School, recently won a national science competition and a $50,000 scholarship. She was one of two first-place winners at the Young Epidemiology Scholars (YES) Competition in Washington, D.C., a contest that attracted 562 entrants from across the U.S.

Michelle's project, “Routine Outpatient Testing of Skin Infections for Methicillin-resistant Staphylococcus aureus (MRSA) in High School Athletes,” investigated whether it is cost effective to require testing of all high school athletes who have skin infections for MRSA. MRSA became a newsworthy disease in 2007 following a number of high school student deaths. A highly antibiotic-resistant pathogen, MRSA does not respond to standard treatments for skin infection.

Read more: MIDAS-Affiliated Student Wins Award

PHDL Selected to Host ISSH 2011


We are pleased to announce that the Public Health Dynamics Laboratory has been selected to host the next Institute on Systems Science and Health (ISSH 2011).

ISSH is a weeklong course hosted at a different site each year, and provides an introduction to systems science methodologies (specifically agent-based modeling, system dynamics, and network analysis). ISSH is sponsored by the NIH Office of Behavioral and Social Sciences Research (http://obssr.od.nih.gov/index.aspx) in partnership with the CDC Syndemics Prevention Network (http://www.cdc.gov/syndemics) and a host site with a strong institutional commitment to systems science and health.

Stay tuned to the BSSR-Systems Science Listserv for announcements about the date, application deadlines, and other information on ISSH 2011. In the meantime, please visit the ISSH website http://issh.aed.org/index.html for information on ISSH 2010 (hosted by Columbia University Mailman School of Public Health) and ISSH 2009 (hosted by the University of Michigan Center for Social Epidemiology and Population Health and the Center for the Study of Complex Systems).

NIH Study Models H1N1 Flu Spread

September 21, 2010

Emily Carlson, NIGMS 
This email address is being protected from spambots. You need JavaScript enabled to view it.

As the United States prepares for the upcoming flu season, a group of researchers supported by the National Institutes of Health continues to model how H1N1 may spread.

The work is part of an effort, called the Models of Infectious Disease Agent Study (MIDAS), to develop computational models for conducting virtual experiments of how emerging pathogens could spread with and without interventions. The study involves more than 50 scientists with expertise in epidemiology, infectious diseases, computational biology, statistics, social sciences, physics, computer sciences and informatics.

As soon as the first cases of H1N1 infections were reported in April 2009, MIDAS researchers began gathering data on viral spread and affected populations. This information enabled them to model the potential outcomes of different interventions, including vaccination, treatment with antiviral medications and school closures. The work built upon earlier models the MIDAS scientists developed in response to concerns about a different potentially pandemic influenza strain, H5N1, or avian flu.

“Computational modeling can be a powerful tool for understanding how a disease outbreak is unfolding and predicting the implications of specific public health measures,” said Jeremy M. Berg, Ph.D., director of the National Institute of General Medical Sciences, which supports MIDAS.  “During the H1N1 pandemic, MIDAS scientists applied their models to see what they could do to help in a real situation.”

Because the H1N1 flu strain is still circulating, a MIDAS group based at the University of Washington in Seattle is now studying the impact the virus could have this fall and winter. Its model, which represents the world population, includes information about immunity—how many people are protected by vaccination or prior infection—and the other circulating flu strains. Using the model, the scientists may be able to predict how H1N1 evolves and the possible role of the H3N2 strain, which historically has been the dominant seasonal flu virus. The results also may help forecast the potential effectiveness of the new flu vaccine that includes both the H1N1 and H3N2 viral strains.

Here are key findings from MIDAS’ earlier work on the H1N1 pandemic. For more results and links to the scientific papers, visit http://www.nigms.nih.gov/Initiatives/MIDAS/Publications.htm.

Estimating Severity
To predict the likely severity of H1N1 in the fall and winter months following the initial outbreaks, the MIDAS group led by Marc Lipsitch, D.Phil., of the Harvard School of Public Health in Boston analyzed patient care data from Milwaukee and New York City. The researchers estimated that about 1 in 70 symptomatic people were admitted to the hospital, 1 in 400 required intensive care and 1 in 2,000 died. They predicted H1N1 to be no more and possibly even less severe than the typical seasonal flu strain. The work, which factored in local differences in flu detection and reporting, also showed that it’s possible to make predictions about severity using data from the early stages of an outbreak.

Vaccinating Children
Ira Longini, Ph.D., at the University of Washington in Seattle and his MIDAS colleagues developed a simulation model to evaluate the effectiveness of different strategies to vaccinate school-aged children, who are known to play a key role in transmitting the flu virus. They modeled a range of scenarios that varied the type of vaccine, the percentage of children vaccinated and the infectiousness of the virus. For each situation, the modeling results indicated that vaccinating this age group substantially reduced overall disease spread and prevented up to 100 million additional cases in the general population. These effects, however, were less strong when the virus was more contagious or when fewer children were vaccinated. Based on these results, Longini’s group concluded that vaccine distribution strategies should depend on a number of factors, including vaccine availability and viral transmission rates.

Cost-Benefit of Employee Vaccination Programs
In one of the first analyses of the economic value of work-sponsored seasonal and pandemic flu vaccine programs, the MIDAS group led by Donald Burke, M.D., at the University of Pittsburgh developed a model that estimated the employer cost to be less than $35 per vaccinated employee with a potential savings of $15 to $1,494 per employee, depending on the infectiousness of the virus.

Interventions and Local Demographics
To determine if a vaccination strategy would likely have the same effect in different locations, a team led by MIDAS investigator Stephen Eubank, Ph.D., of the Virginia Bioinformatics Institute at Virginia Tech in Blacksburg developed models representing the demographics of Miami, Seattle and each county in Washington. The models indicated that while vaccinating school-aged children was the best strategy in each place, the optimal timing and overall effectiveness of the approach varied due to specific characteristics of the local population, such as age, income, household size and social network patterns. These differences, Eubank concluded, suggest that vaccination and probably other intervention strategies should take local demographics into account.

Antiviral Medications
Lipsitch’s collaborators Joseph Wu, Ph.D., and Steven Riley, D.Phil., at the University of Hong Kong used mathematical modeling to predict the likelihood that the H1N1 strain would develop resistance to the widespread use of antiviral medications taken to lessen flu symptoms. Their work showed that giving a secondary antiviral flu drug either prior to or in combination with a primary antiviral could mitigate the emergence of resistant strains in addition to slowing the spread of infection. The results, the researchers concluded, point to the value of stockpiling more than one type of antiviral drug.

School Closures
A public health measure under consideration was closing schools, which previous MIDAS pandemic flu models identified as a potentially effective intervention. According to Burke’s model of Allegheny County, Penn., closing individual schools after they identified cases may work as well as closing entire school systems. When strictly maintained for at least 8 weeks, both types of school closure could delay the epidemic peak by up to 1 week, allowing additional time to develop and implement other interventions. However, the model also indicated that school closures lasting less than 2 weeks could actually facilitate flu spread by returning susceptible students to school in the middle of an outbreak.

“Models like the ones MIDAS has developed help us understand not only trends in disease spread, but also how different factors can influence those trends,” said Irene A. Eckstrand, Ph.D., who directs the MIDAS program. “MIDAS research is leading to new tools and approaches that can aid in making public health decisions at a range of levels, from local to national.”


To learn more about MIDAS, visit http://www.nigms.nih.gov/Initiatives/MIDAS/. To arrange an interview with NIGMS Director Jeremy M. Berg, Ph.D.; MIDAS Director Irene A. Eckstrand, Ph.D.; or MIDAS scientists, contact the NIGMS Office of Communications and Public Liaison at 301-496-7301.

NIGMS is a part of NIH that supports basic research to increase our understanding of life processes and lay the foundation for advances in disease diagnosis, treatment and prevention. For more information on the Institute's research and training programs, see http://www.nigms.nih.gov.

The National Institutes of Health (NIH)—The Nation's Medical Research Agency—includes 27 Institutes and Centers and is a component of the U.S. Department of Health and Human Services. It is the primary Federal agency for conducting and supporting basic, clinical, and translational medical research, and it investigates the causes, treatments, and cures for both common and rare diseases. For more information about NIH and its programs, visit www.nih.gov.

You are here: Home Resources News Archives