Bringing Big Data to Social Science

Mason public policy professor Anne L. Washington uses computational methods to study how government functions could be improved with knowledge management.

Machine Learning Addresses Medical Needs

Mason computer scientist Janusz Wojtusiak says “machine learning” will help computers aid doctors in determining medical treatments, preparing medical bills, and predicting patients’ disabilities.

Researcher Unlocks the Big Potential of Big Data

Professor of astrophysics Kirk Borne explains the logic behind big data, and how it’s essentially changing the way we conduct science.

Using Algorithms to Conduct Large-Scale Metagenome Analysis

Computer Scientist Huzefa Rangwala is developing new algorithms to analyze the functions and relationships of microbial communities.

Kirk Borne

When Disneyland was first constructed, Walt Disney told the builders not to pave the sidewalks. Let’s first see where the people walk, he said.

This logic is behind the concept of data science—looking for the hidden, unexpected patterns in the information first instead of trying to guess what the outcome might be.

Data science, or “big data,” is fundamentally changing the way we conduct science. And that is what Kirk Borne finds so exciting about his job.

“In traditional science, you create a hypothesis—you speculate what you think might happen, and then you set off to prove yourself right or wrong,” says Borne, professor of astrophysics and computational sciences at Mason. “Now, in this age where we have so much data, we don’t have to guess. We just have to sift through all the information and find the patterns.”

These patterns are possibilities and connections that would be impossible to guess without big data. Though the algorithms and mathematics behind data mining are very complex, the concept itself is quite simple. If you collect enough information about something—in the same way a forensic investigator can collect enough information about a crime scene—then you can put a complete picture together.

“There are two things that are great about big data,” Borne says. “You can have the best statistical analysis ever of normal things and also the ability to discover the unusual.”

Model for the Large Synoptic Survey Telescope

Borne first became intrigued with data science more than a dozen years ago when he was working for NASA. Back then, as now, he was considered one of the leading experts on the subject and was asked to brief President George W. Bush on the possibilities, although other circumstances prevented that briefing from happening. “After 9/11, the government was very interested in learning how to look for red flags to try to prevent something like that from happening again,” Borne says.

Since then, big data have become more and more essential to business, government, and science. And despite privacy concerns and worries about what all this information means, the pros outweigh the cons in many people’s minds. And in a world of information sharing through social networks, it is almost impossible to hide from data tracking and still function normally in society.

Log into your Netflix account and all your recommended movies are based on a data science cluster algorithm that analyzes the renting patterns of other people who like the movie you watched last week. Visit a doctor while traveling and electronic medical records will enable you to receive better treatment. Check out at the grocery store and the coupons you receive are targeted to your past transactions.

And for scientists like Borne, data science may lead to the discovery of a new physical law or process. It means revealing new planets, galaxies, and objects entirely.

“My favorite part of data science is discovering the outlier—the thing that doesn’t fit,” he says. Borne calls this “surprise discovery.” Think of the Sesame Street song “One of These Things Is Not Like the Other”—but instead of four Muppets from the neighborhood, Borne is looking at billions of objects in the universe.

“In the era of big data, we’re going to start to see the one-in-quintillion thing that you could never have imagined to discover—the needle in the haystack (if the haystack was the size of Earth). And before, if you did find it, you might think it’s an anomaly—but if you find a bunch of them you can begin to see its reality, begin to see the pattern emerge,” he says.

Borne is working on the Large Synoptic Survey Telescope (LSST) project, a powerful telescope that researchers hope will be built and online by the end of the decade. The LSST will create a 10-year movie of the section of sky visible from its perch atop a mountain in South America.

Borne will help design the data mining techniques that will sift through all of the massive amounts of data gathered and analyzed by the telescope. He is also leading a nationwide scientific collaboration group that will conduct data science research with the LSST data repository, which will be one of the largest scientific databases ever assembled. The LSST data archive will consist of nearly 100 petabytes of data, roughly equivalent to 100 times all the words printed in all the books in all the libraries in the world.

“There are not enough graduate students in the world to look at all those,” he jokes. “So the science part is finding the best algorithms to help make the discovery. The LSST project team will provide open public access to all of these data–it will be the telescope for everyone. The scientific knowledge discovery potential of the LSST database is staggering. And I cannot wait.”

Mason Students Develop Mobile Apps to Discourage Underage Alcohol Use

The Century Council partners with Mason professor Chris Totten to design a mobile game app that conveys the dangers of underage drinking.

Researcher Makes Complicated Big Data Simple to Use

Geoinformation scientist Phil Yang designs systems that help computers organize and interpret information, so it can be readily used by all.

Mason Professor Uses Colorized Computer Models to Evaluate Aneurysms

Colorful computer models created by Mason researcher Juan Cebral give surgeons a whole new way to view brain aneurysms.

The Forecast Is… Murky: The Uncertainties of Weather Prediction

Climate expert Zafer Boybeyi uses advanced numerical models to predict climate and weather events and describes why such work remains challenging.

Chaowei (Phil) Yang

The next time you are enlarging a Google map or clicking on some other global positioning system technology, know that you have researchers such as Mason’s Chaowei (Phil) Yang to thank for such a convenience.

Yang is an architect of sorts, but instead of designing buildings, he is designing systems and creating algorithms that help computers organize and interpret big data in such a way that it is usable by others, especially the general public. In fact, one of his four patents involves the algorithm used to refresh and reposition online maps quickly as the user enlarges or shrinks it.

“We are taking data sets created by others and using them to create new tools that will be beneficial to industry,” says Yang. The list of “others” that work with Yang and his colleagues is long and impressive, incorporating many U.S. agencies, as well as international organizations.

Yang’s work focuses on using spatiotemporal principles, meaning time and space, to optimize computing infrastructure to support scientific discoveries and has been funded by such organizations as NASA, the National Science Foundation (NSF), and the Federal Geographic Data Committee, among others.

“What I enjoy the most [about this type of work] is when a big or complex problem is finally solved through hard work and collaboration,” he says. “Working with others also places us in a better position to solve more problems.”

Since 2010, Yang has served as the chief architect on two NASA projects: Spatial Cloud Computing and Climate@Home. Spatial Cloud Computing is designed to build Cloud Computing platforms using spatiotemporal patterns of data, users, and problems to support NASA science discoveries and engineering development. Climate@Home is an initiative by NASA to create a virtual supercomputer to model climate. Not only are all of NASA’s 10 centers involved, but so are 13 federal agencies and numerous universities and organizations.

Developing the architecture for such a large-scale project isn’t as simple as building it. “We also have to make sure they can maintain the system, and we help them do pilot studies to test it,” he says.

A example from the SilvaCarbon website created by Mason researcher Chaowei Yang’s graduate students.

Here at Mason, Yang directs, with Mason colleague Paul Houser, the NASA/Mason Joint Center of Intelligent Spatial Computing (CISC) for Water/Energy Sciences. Simply put, intelligent spatial computing provides a variety of ways for scientists and other researchers to visualize and manipulate large amounts of data, saving time and resources and ultimately making discoveries and advances in the sciences more accessible. Therefore, Yang and the scientists and graduate students who work with him always have the end users in mind when developing a product or portal.

The joint center was established to respond to community needs and provide leading research and training on intelligent spatial computing internationally. In addition to the center’s partners within the United States, including the National Oceanic and Atmospheric Administration and the Environmental Protection Agency, the center has partnered with a number of Chinese universities and the Heilongjiang Bureau of Surveying and Mapping, the National Geomatics Center of China, and the National Administration of Surveying, Mapping, and GeoInformation.

The center is extraordinarily busy. Yang estimates that there are approximately 30 projects going on at any time.

Among the projects completed by CISC is the very user-friendly SilvaCarbon web portal, which was mainly built by Mason students, according to Yang. The site helps pull together data—including satellite images—from many sources to help countries monitor their forest and terrestrial carbon. Although it is targeted to the world, support for the site comes from a host of U.S. agencies, including USAID and the Smithsonian Institution.

Last summer, Mason and CISC hosted a planning meeting for NSF’s Industry and University Cooperative Research (I/UCR) for spatiotemporal thinking, computing, and applications, which brought together about 50 participants from government agencies, companies, and associations to discuss establishing an industry/university cooperative research center. The cooperative will span three universities—Mason, Harvard, and University of California-Santa Barbara (UCSB)—conduct innovation-focused and deliverable-oriented research with the goal of developing an infrastructure base for future scientific discoveries and engineering development.

According to Yang, this NSF cooperative places Mason as the domain leader for the next decades to come for spatiotemporal studies with impact to human knowledge and problem solving in the 21st century. “We are looking at the next generation [of this field],” says Yang. “This first phase will take place over the next five years.”

Collaborating with the top agencies in the world also has its advantages. “Our collaboration with NASA puts us in the national/international leadership position for geospatial computing and data handling,” he says. “The NASG collaboration puts Mason’s name across all provinces in China, which only a few universities have achieved.”

Ancha Baranova, Konstantin Shakhbozov, and Mikhail Skoblov
RET Finger Protein 2 (RFP2) Promoter

Kenneth J. Hintz
Information Request Generator

Qiliang Li, Dimitris E. Ioannou, Yang Yang, and Xiaoxiao Zhu
Nanowire Field Effect Junction Diode

Siddarth Sundaresan, Albert V. Davydov, Yong Lai Tian, and Mulpuri V. Rao
Microwave Heating for Semiconductor Nanostructure Fabrication

Mark P. Krekeler, Danielle Stoll, Stephen Elmore, and Cynthia Tselepis (Loertscher)
Secondary Process for Radioactive Chloride Deweaponization and Storage

Tariq A. Alsheddi and Ancha Baranova
Method of Identifying Unique Target Sequence

Deapesh Misra
Image Based Turing Test

Ravinderpal S. Sandhu and Joon S. Park
Secure Cookies

Eswar Prasad Ramachandran Iyer and Daniel Cox
Tissue Sample Preprocessing Methods and Devices

Shen-Shyang Ho and Harry Wechsler
Data Stream Change Detector

Harry Wechsler, Hung Lai, and Venkatesh Ramanathan
Recognition by Parts Using Adaptive and Robust Correlation Filters

Sushil Jajodia
Protecting Sensitive Data Associations

Ibrahim S. Abdullah and Daniel A. Menasce
Meta-Protocol

Ruggero Scorcioni, Giorgio A. Ascoli, and Sridevi Polavaram
Neuronal Measurement Tool

Chaowei Yang, Ying Cao, Jibo Xie, and Bin Zhou
Near Real-time Traffic Routing

Xinyuan Wang and Songqing Chen
Transparent Authentication of Continuous Data Streams

Farrokh Alemi
Tailoring Medication to Individual Characteristics

Xinyuan Wang and Daniel Ramsbrock
Live Botmaster Traceback

Sushil Jajodia, Steven E. Noel, and Eric B. Robertson
Intrusion Event Correlation System

Giorgio A. Ascoli and Alexei Samsonovich
A Semantic Cognitive Map

Harry Wechsler and Fayin Li
Face Authentication Using Recognition-by-Parts, Boosting, and Transduction

Babak Jeddi
Methods and Device for Landing Aircraft

Bijan Jabbari, Esmael Dinan, and Rajiv Papneja
Label Switched Packet Transfer Device

Ancha Baranova and Amanda Zirzow
Nanogenomics for Medicine: siRNA Engineering

Brian L. Mark and Ahmed Nasif
Wireless Communications Using Frequency Agile Radio

Yih Huang, Arun K. Sood, and David Arsenault
Self-Cleansing Secure DNS Server

Huzefa Rangwala is a problem solver; no, make that a big problem solver.

Huzefa Rangwala

As a computer scientist, Rangwala is hungry for data on which to test his algorithms. And metagenomics—the collective genome of communities of microbes—provides lots of data.

“If you ask a computer scientist what a genome is, he’ll say it’s a long, long string of characters—it’s a big sequence. Computer scientists get very excited about these kinds of structures,” he explains.

In 2007, the National Institutes of Health launched the Human Microbiome project to study the microbial communities that live within the body and even on a person’s skin: bacteria, fungi, and viruses. More recently, the Earth Microbiome project was established to analyze the microbial communities that live on the planet. Understanding the functions and relationships of these microbial communities to one another—and to their hosts—begins with sequencing the DNA of these communities or reading their structure. But they are so numerous that innovative and scalable computational algorithms must be developed to do it. This is where Rangwala comes in.

As an undergraduate in his native India, Rangwala was trained strictly in computer science. But when he took a bioinformatics class at the University of Minnesota while studying for his PhD, he says, “I got more and more interested in understanding the biology and developing algorithms or methods that might be applicable to biologists.”

Shortly after joining Mason in 2008, Rangwala connected with Mason environmental biologist Patrick Gillevet, director of the Microbiome Analysis Center. A mutually fruitful collaboration developed, and Rangwala has collaborated on several joint projects.

Rangwala says, “Pat has all these biological questions with big challenges, and I’m here to solve those for him. Metagenome association is one of the projects he’s working on related to the field of data mining—finding patterns in data.” “Huzefa has been a very valuable colleague and collaborator,” Gillevet says. “He has helped develop novel algorithms and tools to analyze the microbiome data.”

In developing algorithms to analyze such data, Rangwala aims for speed, efficiency, and accuracy.

“If your algorithm is faster than someone else’s algorithm, then you’ll be able to process the data much faster. We want to split the algorithm and run this algorithm on several machines. Doing that can be challenging because different parts of the algorithm may need to finish at the same time or they might have some sort of dependencies on each other. So, when I’m designing my algorithms, I’m thinking, what is the best way to find concurrency?”

Above that, Rangwala wants to develop algorithms that are user friendly.

When he first began working with Gillevet, Rangwala developed a laboratory information management system for him that included a web interface.

Gillevet’s lab has a sequencing machine that produces about 100,000 sequences of data per run. “How do you store this data efficiently, how do you back it up, how do you transfer it over your Internet or cable? All these factors become core issues, and he was facing these kinds of problems,” Rangwala says.

With the management system in place, the two were able to collaborate more efficiently. “He’s providing me some biological expertise and data, and I’m providing him these tools within a web interface that are really helpful to him. He can proceed with his analysis on his own and repeat them as often as he likes,” Rangwala says.

“He also wanted to analyze these data sets individually on a single machine,” Rangwala continues. “But that takes too long. So we came up with some ideas on approximations. Instead of analyzing the entire set, could we cluster them?” A recent paper Rangwala published with colleagues explained how they developed a process to do that: putting similar examples or similar objects in the same groups and analyzing the representatives for each group.

With the technology available, Rangwala can sequence all the bacteria on a slide sample. But since all the bacteria are mixed together, the sequence doesn’t tell biologists what they need to know: what kinds of bacteria there are, how abundant they are, which are the dominant species, and what the bacteria do.

“That’s a huge problem,” Rangwala says. He’s now working on approaches that will extract the underlying relationship between these different problems—combine them and produce a better annotation of the bacterial species.

In addition to working with Gillevet, Rangwala is collaborating with other researchers at Mason and Rush University Medical Center in Chicago, the University of Minnesota, and the U.S. Department of Agriculture (USDA). His research is supported by grants from Mason, the National Institutes of Health, the National Science Foundation, the Defense Advanced Research Projects Agency, and the USDA.

“My job is to create new algorithms, but I’m excited that [my work] has an impact in the fields of biology and environmental sciences right now,” Rangwala says.

Predicting weather and climate change has long been a delicate process, and Mason’s Zafer Boybeyi is the first to admit that even with today’s technology, certain aspects of the science remain largely unpredictable.

Zafer Boybeyi

An associate professor in Mason’s Department of Atmospheric, Oceanic, and Earth Sciences for the past 15 years, Boybeyi is also director of the Comprehensive Atmospheric Modeling Program (CAMP) within the College of Science. The program, formed in 1997, originally received funding from the Department of Defense to develop emergency response systems for chemical, biological, nuclear, and radiological releases. Over the years, that research has expanded to encompass climate studies, aerosol impacts, severe weather, and natural hazards.

Modeling and Simulation

Numerical modeling—using mathematical models of the atmosphere and oceans to predict weather—has come a long way from efforts that started nearly five decades ago with 1-D and 2-D simulations. Boybeyi says today’s prediction models, particularly those based on the weather scales of a few days, are exceedingly reliable, mainly because of advanced computers and denser observational networks. These networks include conventional systems such as rawinsondes—instruments that measure wind, pressure, temperature, and humidity—and unconventional ones such as radar and satellites.

But weather prediction is quite different from foretelling climate patterns because, says Boybeyi, “you’re making much longer predictions—10, 20, 50 years down the road. The accuracy of those models deviates from the reality as time goes by.”

Researchers are really looking for trends in these model predictions. If a model shows both carbon dioxide levels and temperatures increasing consistently, for example, then a strong correlation must exist between the two. Numerical models are used to interpret those relationships.

“And as we predicted,” Boybeyi says, “when you increase greenhouse gases, you increase global temperature. There’s a strong correlation there, and there’s no doubt about it. How much temperature change will take place? No one has a clear answer for that.”

A Bumpy Ride

In the beginning, modeling presented a learning curve for researchers. But as numerical models became sophisticated, research caught up, reached a point, and then leveled off. Little progress is being made now, mainly because of turbulence, what Boybeyi calls “the one classic, unsolved physics problem.”

Scientists currently have no way of mathematically defining the random motions involved in turbulence because the atmosphere is both deterministic (i.e., predictable) and chaotic in nature. Therefore, Boybeyi says, “there are both theoretical and practical constraints on what we can do. Think about atmosphere as so many parameters affecting our temperature, density, pressure, and all the weather events that we care about. One parameter affects so many other parameters, and so many other parameters affect others. These nonlinear interactions and relationships are what we fail to understand clearly.”

But in the future, he says, computer advancements will enable modeling to produce higher resolution simulations that will reduce the problem of mathematically defining these random motions, and “resolve explicitly most of the small motions and the nonlinear interactions that exist in the atmosphere.”

Predicting severe weather events also presents an ambiguous challenge. Hurricanes, for instance, have been studied for such a long time, there is now a high level of success in predicting their tracks. But as for predicting their strength, Boybeyi says, “We don’t really do a good job on that. We fail to predict the intensity with accuracy.”

Boybeyi’s group is examining how the Saharan Air Layer (SAL) affects hurricane intensity. The SAL is a mass of dry, dusty air that forms over the Sahara Desert during spring, summer, and fall, and then moves out over the tropical North Atlantic Ocean.

“A recent hypothesis suggests that the Saharan Air Layer, and the dust it carries with it, may cause the intensity prediction problem in the Atlantic region,” Boybeyi says. “That’s a relatively new idea, and people have just started paying attention to that.”

Aerosol Effects

These days, Boybeyi’s group is putting nearly all its efforts into studying the impact of aerosols on climate. Aerosols, or all the particles dispensed in the air including greenhouse gases, have different impacts in the atmosphere. Direct impact, in which aerosols block, scatter, and absorb the short-wave radiation coming from the sun, reduces the amount of solar radiation reaching the ground’s surface and alters the entire energy balance of the atmosphere. Indirect impact occurs when aerosols provide cloud condensation nuclei. This increases the lifetime and thickness of clouds, which delays precipitation. As a result, those clouds can block short-wave radiation and trap long-wave terrestrial radiation.

Boybeyi’s team has also been working on a third, semidirect effect caused by highly absorbing aerosols, such as black carbon. This dark soot soaks up sunlight and generates heat, and aerosols of this type, he says, can “absorb energy and radiation, and increase the atmospheric temperature further. They may play a tremendous role.”

Some of these aerosol impacts contribute to global temperatures in a positive way, and some in a negative way. Which does which is another uncertainty. “Researchers as yet have no clear understanding of their total contribution and which one is the dominant process,” Boybeyi says. “But the polar regions may hold the answers.”

As the earth’s general air circulation pulls in greenhouse gases released from industrialized regions of the world—particularly developing areas such as Asia and parts of Europe—it transports them to the polar regions, where they combine and increase their concentration level. One of Boybeyi’s graduate students, Eric Stofferahn, has been working on aerosol effects on climate in the Alaskan region, through research sponsored by the Department of Energy.

“Our preliminary results indicate that these aerosols play an important role in terms of the vast climate change happening in the polar regions—that of the icebergs melting,” says Boybeyi. “The models predict that in 10 to 20 years, there may be little ice left in those regions during summer months.”

Finally, Boybeyi says long-term, sustainable development of our global society primarily requires an understanding of the interaction between human activities and natural processes (such as climate and climate change), and accurate representation of these natural processes using numerical models. “Unfortunately,” he says, “this remains a challenging problem.”

Juan Cebral

Juan Cebral’s complex computer model does more than simply show blood swirling in a brain aneurysm’s labyrinthian pattern; it helps doctors determine whether the aneurysm is about to rupture and needs surgery.

Cebral, a professor at Mason’s Center for Computational Fluid Dynamics in the College of Science, studies fluids and how they move. He works with surgeons and other researchers to map out how blood flows in the brain, specifically in aneurysms. Brain aneurysms happen when blood bulges against a weak arterial wall and creates what resembles a balloon in the artery. It’s unknown why aneurysms happen.

Cebral is working with Inova Fairfax Hospital; the University of California, Los Angeles; the Mayo Clinic; the Buenos Aires, Argentina-based Clinical Institute ENERI; and Philips Healthcare in the Netherlands.

“I feel very excited that maybe my work will help doctors or patients,” says Cebral. “If you look at the history of medical advances, doctors and scientists working together have made these advances. The doctors don’t have the time to do research most of the time. The scientists need patients to do the research. It’s difficult to get the right combination.”

Technology is catching up with the complexity of the human brain and is making it possible to study the interaction between blood flow and neurons. Past models were idealized, says Cebral, who earned his doctorate in computational fluid dynamics from Mason in 1996. “For instance, when you build a glass model of an aneurysm, it’s just a straight glass tube with an aneurysm. And that never happens in the human body.”

But now computerized models are a wonder of colors, and Cebral’s models give surgeons a picture of the aneurysm that they wouldn’t have had otherwise. He starts with the gray x-ray image of a patient’s aneurysm and transforms it into swirling colors to show the complex blood flow. Blue is normal blood flow, while red shows problematic blood flow. It can take a day or two to build a complex model and then another day to run the simulation.

“You can use the arterial geometry of specific patients and then you can simulate what that blood flow looks like,” Cebral says. “You can study specific flow dynamics. We can provide some information that the doctors don’t have today, which is the fluid dynamics of each individual patient.”

Mason researcher Juan Cebral present on the different types of aneurysms he has analyzed at the Center for Computational Fluid Dynamics. Photo by Creative Services.

About 5 to 8 percent of people have a brain aneurysm, but less than 1 percent rupture every year, Cebral says. Intervention is risky; it may not be worth taking that risk if the aneurysm isn’t going to rupture.

“So the first question we need to ask is, which aneurysm is most likely to rupture?” Cebral says. His computer models work to answer that question by measuring the force of blood on the arterial wall and compiling statistics of the results.

“With that information we look at what happened with these aneurysms. Did they rupture? Did they not rupture? We look for the conditions in ruptured aneurysms and how they are different from the conditions in unruptured aneurysms. If we find that, for instance, an unruptured aneurysm has smoother, simpler blood flow than a ruptured aneurysm, we can use that information to try to predict whether an aneurysm will rupture.

“So if someone comes into the hospital and we see pressure on the arterial wall, then we can say, ‘I’ve seen these characteristics on most of the ruptured aneurysms, so this is more likely to rupture.’”

That brings treatment into the picture. Cebral is working with Ram Kadirvel, an assistant professor of radiology at the Mayo Clinic, on a $2 million grant from the National Institutes of Health to study a new treatment. Work started in late spring 2012, and Kadirvel credits Cebral as the main reason they received the grant.

Currently, surgeons use platinum coils inserted into the aneurysm to avoid open skull surgery. But this option doesn’t last, and patients need to be re-treated, Kadirvel says. They’re testing whether a stent—a mesh metallic device placed across the aneurysm’s neck—could offer a better option. They’re also studying high-flow, medium-flow, and low-flow density stents to see which offer the best chance for healing the aneurysm.

“Juan puts all the data together and then he creates a model for before and after the stent,” Kadirvel says. The stents could be used in humans in four to five years.

Cebral says more options could be possible as researchers are able to study aneurysms in greater detail. In 2005, he had 60 aneurysm models taken from patients, and that was top of the line. Today, he can compare as many as 200 patients.

Fluid dynamics also can be applied to the heart, legs, and other arteries. “There are many fluid mechanics problems in the human body,” Cebral says.

Janusz Wojtusiak

By analyzing huge swaths of patient data, computers are learning not only how to help doctors choose the best medical treatment for each patient, but also how to efficiently prepare medical bills and predict patients’ disabilities.

Dubbed “machine learning,” complex computer algorithms delve into data to boost individualized medicine, says Mason computational scientist Janusz Wojtusiak, director of the Machine Learning and Inference Laboratory and the Center for Discovery Science and Health Informatics at Mason’s College of Health and Human Services.

“Traditional research and traditional clinical trials focus on an average patient,” Wojtusiak says. “But if you’re a patient, you don’t care about the average patient. You want one specific treatment that will benefit you, have minimum risk, and will have the highest chance of a good outcome for you.”

Machine learning can help find the answer, Wojtusiak says. “It can be used to build individualized models. By observing patterns in past patients, machine learning can find out what are the best options for you, not an average.”

Complex algorithms help computers learn by doing, much like a child does with language or tasks, Wojtusiak says. The approach is similar to Amazon and Netflix when books and movies are recommended based on past choices.

The goal isn’t to replace the expertise of physicians but to give them the most complete information so they can make the best decision, Wojtusiak says. Inches-thick patient files are transformed into compact data. Lab tests, doctor’s notes, demographic information, and other details build an elaborate picture of that patient. When those data are combined with information about other patients, the computer can find new information that’s more than just numbers.

“With many machine learning methods, people analyze huge amounts of relatively simple data,” Wojtusiak says. “In health care, these are not just numbers, these are specific lab tests, specific diagnosis, specific treatments, specific notes that describe real patients. The true challenge is to create smart algorithms that understand what the data actually mean and put some meaning behind the data, and then be able to learn something better and easier from it.

Computers do more than create electronic records. Computers also can change the way patients live. One of the ways Wojtusiak’s team is doing this is through a pilot program with the Department of Veterans Affairs (VA).

The VA recently launched a two-year pilot program in Bay Pines, Florida, to help 45 veterans, who typically would live in nursing homes, stay in regular homes. These “foster” homes frequently are with family members, and the VA provides the medical care at these residences and requires they be extensively evaluated before they are approved as foster homes.

“[The pilot] started as an initiative to allow patients to stay where they like at a home instead of at an institution,” says Farrokh Alemi, an affiliate researcher at Mason’s Center for Discovery Science and Health Informatics and chief of performance improvement for the VA.

To assess how foster home patients are doing compared with those in nursing homes, the VA will ask questions to determine whether, for instance, patients fall more or less often or whether they are more or less depressed.

Alemi says they are using that information to evaluate foster home care as an alternative to nursing home care. “We’re looking at the outcomes of this care.”

Wojtusiak’s machines are crunching data that could be used to open up the program to 800 patients, Alemi says. “If the data support this, there will be rapid expansion of the program.”

Researchers at the Machine Learning and Inference Lab are also studying prostate cancer patients, comparing five different treatment options.

“The idea here is to find out which patients are similar, and then for similar patients, to find out what are the best treatment options,” Wojtusiak says. “Then when a new patient comes in, we can find out whether the patient is similar to those we analyzed before and whether specific results of comparison are helpful.”

In September 2012, Wojtusiak began work on another project with the VA that examines 2,000 heart failure patients. He projected that millions of lab tests and physician notes will be examined over the course of the study.

“We can identify what works and when it doesn’t work,” Alemi says. But machine learning can take the research further than that.

“We can predict who will live and who will die with great accuracy,” Alemi says. “It makes my hair stand on end. It’s an amazing time to be in health care.”

Despite the impact, people aren’t going to do what a computer advises unless they understand the science behind it. And it’s not enough for the algorithms to be accurate. The data need to be explained if they are to be accepted by physicians, according to Wojtusiak.

“It’s not a black box,” Wojtusiak says, comparing machine learning to the aviation device.

You can imagine that if you went to a surgeon and told him the “black box” suggests he operate on a patient, he or she is not likely to do it, says Wojtusiak. But if the computer model is able to show exact evidence behind the suggestion—based on specific patient factors and specific patterns based on previous patients—that the best course of action is to perform a specific type of a surgery, then the doctor is more likely to act on it.

In the end, all the number crunching is worth it. “Health care is the perfect area for machine learning methods because it’s extremely complex on one side and it’s hard to deal with on the other side,” Wojtusiak says. “You can really help people.”

Ancha Baranova

Gene therapy and its potential to further personalized medicine is one of the hot topics in medicine today. Gene therapy derives its name from the idea that DNA can be used to alter or correct part of the genetic strand, affecting an individual’s cells to treat or prevent disease.

One of the challenges facing doctors and researchers pursuing this line of inquiry is how to deliver the therapeutic small interfering RNA (siRNA) to the appropriate site in the body. Mason biologist Ancha Baranova’s latest patent provides such a delivery system.

Oral delivery doesn’t work for this kind of therapy, Baranova says, because the siRNA could not survive the digestive track and its acids. Intravenous delivery with an injection into the circulatory system presents other obstacles.

“To the body, [the siRNA] looks like the sign of a viral infection so it can activate an immune response,” she says. This response can also lead to the siRNA’s degradation.

To thwart the body’s immune response, Baranova has developed a cloaking device to prolong the life of the siRNA and improve its chances of reaching the target site. The therapeutic molecules are tucked into what Baranova calls a “DNA basket.”

“You can do amazing things with DNA,” she says. This basket is made using DNA from salmon, which is cheap and clean, according to Baranova. In the future, she believes the patient’s own DNA could be used to create the basket, ensuring even greater compatibility.

Baranova shares the patent, Nanogenomics for Medicine: siRNA Engineering, with her former graduate student Amanda Zirzow. It is the fourth patent that Baranova has been awarded since joining Mason in 2001.