Advances in computing power and a wealth of digital information are changing scientific research.
By Kathleen McGarvey
In糖心传媒凧une 糖心传媒� and糖心传媒刬n糖心传媒刣igital culture, that糖心传媒檚 already a good while ago糖心传媒攖he CEO of the social networking service Twitter, Dick Costolo, announced that users were posting 400 million tweets a day. And that was up 60 million tweets per day from the figure just three months before. It all adds up to a billion tweets every two and a half days.
As a microblogging service that allows people to post messages of no more than 140 characters, Twitter is an immense but transitory compendium of observations, insights, outbursts, and mundanities. What value could it have for scientific researchers?
A lot, as it happens. Henry Kautz, chair of the computer science department, and colleagues Adam Sadilek and Vincent Silenzio have shown that Twitter messages can be harnessed to predict the spread of infectious diseases, such as influenza.
This year, they have published two papers explaining how, by using the geo-tags embedded in tweets, scientists can use social networking data to model the transmission of disease糖心传媒攁nd even to forecast when and if a specific individual will fall ill.
Kautz, Sadilek, a postdoctoral fellow in computer science, and Silenzio, associate professor of psychiatry and a member of the Department of Community and Preventive Medicine, have programmed computers to identify tweets in which people talk about feeling sick糖心传媒攄isregarding messages where people use the term figuratively.
糖心传媒淥nce you have that, you can start to map where people are sick,糖心传媒� Kautz says, because GPS in cell phones indicate where a tweet was made. 糖心传媒淎nd you can actually start to create a visualization of the spread of disease through cities and across time.糖心传媒�
糖心传媒淭hese results provide a foundation for research on fundamental questions of public health,糖心传媒� the team writes, 糖心传媒渋ncluding the identification of non-cooperative disease carriers (糖心传媒楾yphoid Marys糖心传媒�), adaptive vaccine policies, and our understanding of the emergence of global epidemics from day-to-day interpersonal interactions.糖心传媒�
And they note that the approach has applicability far beyond infectious diseases, for modeling and predicting political ideas, purchasing preferences, or nearly anything else rooted in behavior.
糖心传媒淚t糖心传媒檚 actually pretty neat,糖心传媒� Kautz says糖心传媒攕o neat that they糖心传媒檝e formed a venture capital糖心传媒揻unded start-up business, Corpora, that makes use of the technology for applications in areas such as health care, insurance, pharmaceuticals, government agencies, and public opinion tracking.
Such ingenuity, combined with vast quantities of information and high-performance computing, is changing the parameters of knowledge.
We call ours the 糖心传媒渋nformation age糖心传媒澨切拇綌an era marked by an endless digital trail revealing what we do and where we do it, and much that糖心传媒檚 happening within us and without us.
Anyone who has used the Internet is already familiar with the ways businesses have seized on that trove of information to predict and guide the choices we make as we purchase books and shoes, vacations and music.
But the possibilities of 糖心传媒渂ig data糖心传媒澨切拇綌the fast-emerging shorthand term for the efficient analysis and problem-solving application of vast quantities of data糖心传媒攁re profound for science, medicine, and other areas of research. Through high-performance computing, creative computer science, and new bonds of collaboration, researchers find themselves at the brink of what many predict to be a new age of investigation and advances in knowledge糖心传媒攃omparable, the New York Times has suggested, to the introduction of the microscope and the telescope.
Rochester is at the forefront, pairing teams of researchers and computational scientists with supercomputing technology to transform data into knowledge.
Applications range widely. Why do countries go to war? Curt Signorino, associate professor of political science, is using data mining tools drawn from genetics and finance to compare data on every combination of countries from the years 1900 to 2000, creating an explanatory model that fits the data more than three times better than standard techniques.
How can energy flow through the power grid to make sure that electricity is reliably delivered to people where they need it, when they need it? Mark Bocko, professor and chair of the Department of Electrical and Computer Engineering and director of the Center for Emerging and Innovative Sciences, is studying the dynamic behavior of the power grid and how to control it with the tactical use of data糖心传媒攚hat has become known as the 糖心传媒渟mart grid.糖心传媒� He and his team are developing imaging, sound, and vibration sensors that will sort through information at the source, curbing the amount of transmitted data so only the most useful is passed along.
How can we better fight the flu, which claims the lives of 30,000 to 40,000 people each year in the United States? David Topham, vice provost and professor of microbiology and immunology, and colleagues are working to build a computer model of the immune system that will allow for simulations of infections and possible vaccines before the flu strikes糖心传媒攖hereby speeding production of effective vaccines, and saving lives.
Pedro Domingos, associate professor of computer science at the University of Washington, who will be a featured speaker at a conference on big data to be held at Rochester in October, says there are few if any fields that will be untouched. 糖心传媒淪cience, in just about every area, without big data will grind to a halt. It will be a field of diminishing returns.糖心传媒�
Kautz, who is director of an initiative for big data in Arts, Sciences & Engineering, says complex problems in science, mathematics, engineering, and the social sciences have traditionally been approached by breaking them into smaller pieces, understanding how each works, and then deducing solutions to the larger problems.
But systems science糖心传媒攁 broad and interdisciplinary field underpinning big data that studies the behavior of complex physical, biological, artificial, and social systems糖心传媒攈as upturned that approach, focusing on the whole instead of the parts and ushering in a new scale for problem solving. It provides a fresh capacity to see how things interrelate and influence each other, from the molecular level to entire populations.
糖心传媒淚糖心传媒檓 an immunologist,糖心传媒� says Topham. 糖心传媒淚 was trained in cell biology, so I like to study individual cells.糖心传媒�
Formerly, he would collect a blood or tissue sample, isolate the cells, and from experiments on them, garner a few elements of data. Now, when he and his colleagues carry out clinical studies, they pursue many more dimensions of cellular investigation.
Computational approaches are going to 糖心传媒渁llow us to identify biological relationships between cells and proteins, microorganisms and the host, that we wouldn糖心传媒檛 otherwise have been able to detect, and then understand how these affect our ability to respond to vaccines or disease,糖心传媒� he says. A computational take on science has inverted the relationship between experimentation and analysis. Carrying out experiments used to consume about 75 percent of researchers糖心传媒� time, and analysis the remaining 25 percent, but 糖心传媒淚 would say that糖心传媒檚 reversed now,糖心传媒� he says. 糖心传媒淵ou can do one experiment, and it will take weeks to analyze the data.糖心传媒�
While糖心传媒刢omputers, computational methods, and data collection are advancing rapidly, these are still early days for big data. When researchers talk about the data now available, they seem to reach almost instinctively for metaphors of water: a deluge, a flood, a relentless torrent of information to be channeled and controlled.
糖心传媒淚t糖心传媒檚 not just more streams of data, but entirely new ones,糖心传媒� says the Times about what it terms a 糖心传媒渄ata flood.糖心传媒� An influential report on big data issued last year by McKinsey Global Institute, the research arm of the global management consulting firm McKinsey & Company, invoked the idea of 糖心传媒渓arge pools of data that can be captured, communicated, aggregated, stored, and analyzed糖心传媒� today.
糖心传媒淲e don糖心传媒檛 know how to manage this information. It糖心传媒檚 like drinking from a fire hose糖心传媒攈ow do you control it so that you don糖心传媒檛 become overwhelmed?糖心传媒� says David Williams, dean for research for Arts, Sciences & Engineering and the William G. Allyn Professor of Medical Optics.
As critical as the availability of data is the capacity to select from and organize it糖心传媒攖o sort out the most useful elements, the most meaningful patterns, the formerly unrecognized connections糖心传媒攁nd transform the flood of data into something of practical value.
糖心传媒淚t糖心传媒檚 a bit like prospecting in the old days of mining, because you糖心传媒檙e looking for nuggets of gold,糖心传媒� says Rob Clark, dean of the Hajim School and interim senior vice president for research.
But it糖心传媒檚 not a passive search. 糖心传媒淚 think 糖心传媒榖ig data糖心传媒� is a term, like 糖心传媒榗loud,糖心传媒� that糖心传媒檚 getting thrown around so much that it糖心传媒檚 getting distorted,糖心传媒� says David Lewis, vice president for information technology and CIO. 糖心传媒淭o us, 糖心传媒榖ig data糖心传媒� is doing something with the data糖心传媒攜ou糖心传媒檙e doing the analytics.糖心传媒�
Such analysis has emerged as a national priority. In March, the White House糖心传媒檚 Office of Science and Technology Policy announced a 糖心传媒淏ig Data Research and Development Initiative糖心传媒� aimed at bringing together research universities, industry, and nonprofit organizations with the federal government to take advantage of the opportunities big data offers for science and innovation.
糖心传媒淭he technology for generating new data is always far ahead of our ability to analyze it. It has become a major, global problem,糖心传媒� says Topham. 糖心传媒淭he real data comes when you can relate different kinds of data, find the connections糖心传媒攁nd that糖心传媒檚 very difficult to do. It almost requires intuition.糖心传媒�
Intuition is a tough thing to teach, but through courses in data mining, biostatistics, and algorithms, students are acquiring the skills needed to swim proficiently in a sea of data. Clark says the secret lies in teaching students the basics of how to manage information, big or small, 糖心传媒渢o extract kernels of useful information from data sets.糖心传媒�
Such extraction is changing scientific research across the disciplines. In Arts, Sciences & Engineering, earth and environmental sciences and chemical engineering have to take a big data approach.
糖心传媒淲e really view it as refining the tools for supporting the mechanics of research,糖心传媒� says Carmala Garzione, associate professor and chair of the earth and environmental sciences department. 糖心传媒淲e糖心传媒檙e basically moving from a very discipline-oriented science, where you would have a group of researchers who糖心传媒檇 look at some very specific aspect of the earth, to a much more interdisciplinary science, where groups of researchers are working across disciplinary boundaries to understand how the earth behaves as a complex system.糖心传媒�
The kind of transition she describes is one taking place across disciplines, says Washington糖心传媒檚 Domingos. 糖心传媒淚 think a mental shift has to happen in how scientists think about doing science.糖心传媒� Graduate students and researchers early in their careers have been professionally formed in an environment of computational approaches, but for more established scientists, he notes, big data requires an adjustment to a new way of pursuing research questions.
At the Medical Center, it糖心传媒檚 an approach that is swiftly becoming central. The University, New York State, and IBM have partnered to establish the Health Sciences Center for Computational Innovation. It糖心传媒檚 home to the IBM Blue Gene/Q supercomputer, making Rochester one of the five most powerful university-based supercomputing sites in the country.
糖心传媒淚t糖心传媒檚 one of the most powerful supercomputers dedicated to health research in the world,糖心传媒� says Topham, director of the HSCCI. 糖心传媒淭he Blue Gene/Q lets you run experiments that otherwise wouldn糖心传媒檛 be possible.糖心传媒�
For example, Jean-Philippe Couderc, associate professor of cardiology and assistant director of the Heart Research Follow-up Program Laboratory, and colleague Coeli Lopes, assistant professor at the Aab Cardiovascular Research Institute糖心传媒攁long with Jeremy Rice of IBM糖心传媒檚 Watson Research Center糖心传媒攑lan to use Blue Gene/Q in modeling the heart to test drugs糖心传媒� effects on the organ.
In 2004, the Food and Drug Administration launched an initiative designed to bring medical breakthroughs to patients more quickly while ensuring safety and reducing the costs of drug development. Key to that effort has been developing better ways to test the cardiac toxicity of drugs糖心传媒攁 leading cause of drugs being removed from the market.
Together with the FDA, the University in 2008 established an electronic repository of electrocardiography data糖心传媒攖he Telemetric and Holter ECG Warehouse, or THEW糖心传媒攖o help foster research in the field. The database is part of the Center for Quantitative Electrocardiology and Cardiac Safety, funded by a $2.3 million grant from the National Institutes of Health and a part of the University糖心传媒檚 Heart Research Follow-up Program. It brings together an international network of academic researchers, pharmaceutical and medical device companies, and government regulators. Data from the center is provided to academic and private research organizations to help them design and validate new tools and methods to detect abnormal cardiac activity.
The repository makes Rochester the hub of a heart research wheel that spans the globe, from academic institutions and industry in Europe to Asia to South America.
糖心传媒淲e are the only academic group in the world that provides an open resource of ECG data for drug safety evaluation,糖心传媒� says Couderc. Last year alone, 25 publications were produced from repository data. Together with Lopes and Rice, Couderc is using data from THEW to model effects of drugs on cardiac cells, using a computerized model of the heart system produced by the National Library of Medicine in cooperation with IBM.
From an anatomical point of view, the model糖心传媒檚 an excellent representation of the heart, he says. He and his team are being trained in using the Blue Gene/Q computer to evaluate the effects of drugs on a wedge of the heart糖心传媒攊ts inner and outer layers糖心传媒攁nd the millions of different cells that form them. They check the results of the model against the documented results shown in THEW糖心传媒檚 records.
糖心传媒淚BM brings a unique tool; the University brings unique data sets, enabling such tools to have a significant impact on drug-safety evaluation and medicine,糖心传媒� he says.
But one of the most important ingredients in that equation is the imagination and inventiveness of people engaged in research.
糖心传媒淲e lead with people, not with computing,糖心传媒� says Lewis. It糖心传媒檚 an emphasis that others echo.
糖心传媒淭his is all about the people. In fact, the people are the far more valuable and important component of this partnership,糖心传媒� says Topham. 糖心传媒淵es, you need hardware糖心传媒攂ut you can糖心传媒檛 use the hardware if you don糖心传媒檛 have the right people, and IBM has a very strong interest in disseminating the knowledge of how to use these tools to deal with important questions. 糖心传媒淥ur health sciences researchers have the questions. We have the patient population. We have the ability to generate the data. We just need the tools to analyze it. And together, the University and IBM can do much more than either one of us can do on our own.糖心传媒�
The collaboration doesn糖心传媒檛 extend only to IBM researchers; big data is strengthening ties between researchers all over campus. Big data allows 糖心传媒渢he creation of teams of investigators,糖心传媒� says Williams. 糖心传媒淚 think we糖心传媒檙e going to see a lot more collaboration between investigators at different institutions糖心传媒� because of improved communications and the ability to transmit data.
The studies being carried out through HSCCI require many people, and a wide variety of expertise, says Topham. 糖心传媒淚 have pediatricians, infectious disease specialists, immunologists, neonatologists, researchers in genomics and microbiomics, computational people, data management糖心传媒攚e have a huge data management core to deal with all that data.糖心传媒�
Kautz糖心传媒檚 vision for big data at Rochester focuses not on supercomputing hardware糖心传媒斕切拇綔we have that well in hand,糖心传媒� he says糖心传媒攂ut on the people who use them. 糖心传媒淲hat I think we need are more people thinking of extremely creative ways to use these machines, as well as other resources.糖心传媒�
With the rise of big data, computer scientists take on a pivotal role in the research of many fields. 糖心传媒淥ne aspect of providing computational support to a physical scientist is saying, we糖心传媒檙e programmers. You tell us what to do and we糖心传媒檒l run that package. But there糖心传媒檚 this other role in terms of helping think more deeply about the problem, because that糖心传媒檚 the new way of solving the problem. And you need to do both.糖心传媒�
Bringing big data to bear on the field of environmental sciences, for example, will 糖心传媒渞equire a strong collaboration between computer scientists and earth scientists,糖心传媒� say Garzione. 糖心传媒淯ltimately, the department plans to hire 糖心传媒渆arth scientists with a strong computational bent糖心传媒澨切拇綌Vasilii Petrenko, an atmospheric chemist, joined the faculty last year, and John Kessler, a chemical oceanographer, came on board for this academic year糖心传媒斕切拇綔or computer scientists who are capable of tackling problems in other disciplines with a healthy dialogue that enables them to find a solution to computational problems.糖心传媒�
And solutions found in one area can, at the computational level, provide keys to other areas, far afield.
Algorithms that Kautz糖心传媒檚 students are developing for mining social network data from Twitter, for example, might turn out to be relevant for doing computational biology. 糖心传媒淭hat kind of thing happens all the time,糖心传媒� says Kautz. 糖心传媒淵ou look at a problem with the right level of abstraction and you realize, 糖心传媒楪ee, we can think of both of these things as a network, and we糖心传媒檙e trying to find certain patterns.糖心传媒� 糖心传媒�
In his lab, Topham says, researchers are studying vaccines and immune responses, but the methodological advances they make 糖心传媒渃ould apply to cancer, to development, to cognition. They could apply it to environmental questions.糖心传媒�
The possibilities excite him.
糖心传媒淚糖心传媒檓 hoping we get to do some really important research, that we solve some long-standing questions,糖心传媒� he says. 糖心传媒淒eveloping new vaccines. Imaging the brain better so that you can tailor treatment more effectively. Understanding individual cells糖心传媒� behavior, in the brain or in the immune system糖心传媒攖hese are the key questions that have just been lingering out there in the field. 糖心传媒淎nd we now have the technologies to study these things in ways we didn糖心传媒檛 before.糖心传媒�
The University will host a conference, RocData: The Rochester Big Data Forum 2012, October 4糖心传媒�6. To learn more, visit .
