What makes content go viral? ASU computer scientist investigates

April 1, 2015

For years Arizona State University computer scientist Paulo Shakarian has been trying to figure out what makes information go “viral” on social media – as well as what keeps it from happening.

One possible answer occurred to him on a visit with his young son to the Hall of Flame Fire Museum in Phoenix, which is dedicated to the historical preservation of fire fighting equipment used around the world. Shakarian viral on social media Download Full Image

“I was reading about an exhibit on the firefighters who fight rural fires by creating firebreaks, and it gave me an idea. Perhaps information spreading in social media is not able to reach viral proportions because it reaches a ‘firebreak,’” said Shakarian, an assistant professor of computer science and engineering in the School of Computing, Informatics, and Decision Systems Engineering, one of ASU’s Ira A. Fulton Schools of Engineering.

The potential firebreak in this instance would be in the form of a group of people who are not well connected to others in a social network, or are part of a community that is simply not interested in the message. Thus, they inhibit information from spreading.

His interest and initial investigations into the area has led Shakarian to recently be awarded a grant of $354,000 from Young Investigator Research Program of the Air Force Office of Scientific Research. His research proposal, titled “Toward Anti-Inhibitory Influence of Online Social Networks,” was one of only 57 selected from more than 200 proposals for funding from the program.

The grant program aids researchers in the early years of their careers whose work demonstrates potential for achieving significant advances in their fields of engineering and science, with the objectives of fostering creative basic research in science and engineering, enhancing early career development of outstanding young researchers and increasing opportunities for the researchers to recognize the Air Force mission and the related challenges in science and engineering.

The Air Force is funding Shakarian's research because of its interest in gaining a better understanding of how information propagates in social media.

The ability to identify social network inhibitors interests researchers, industries and governments in many ways. Marketers, for instance, want to identify the inhibitors so they can take steps to work around them and reinvigorate their campaigns with a new plan. It can also be useful in identifying what inhibits content when, for example, government wants to limit or stop messages being spread by potential terror groups.

Shakarian found that many other scientists noticed different things that act as inhibitors, but the topic was typically discussed only as side notes in research papers. So he began compiling a list of things that might inhibit the spread of information in a social network, and has been working to find a way to navigate around each of these obstacles.

While protecting the privacy of social media users, he and his research team have been looking at anonymous data from Twitter, Instagram and Weibo (a Chinese microblogging site with more than 500 million users). With all of the content created on social media, the percentage that goes viral is very small, offering only a limited amount of examples to work with.

“Very rarely does a post on social media gain traction beyond immediate friends,” Shakarian said. “We want to understand why that is and build algorithms and software to overcome these obstacles.”

Shakarian is a 2002 computer science graduate of the U.S. Military Academy at West Point. He held various military positions with the Army’s First Armored Division and had two deployments to Iraq.

He earned master’s and doctoral degrees in computer science from the University of Maryland, College Park, and then returned to West Point as an assistant professor in 2011, before joining the ASU faculty in 2014.

Written by Erik Wirtanen

Joe Kullman

Science writer, Ira A. Fulton Schools of Engineering


Evolution of blood cancer subject of ASU professor's research

April 2, 2015

All living things evolve over time, and cancer cells are no exception – subject to the two overarching mechanisms described by Charles Darwin: chance mutation and natural selection.

In new research, Carlo Maley and his colleagues describe compulsive evolution and dramatic genetic diversity in cells belonging to one of the most treatment-resistant and lethal forms of blood cancer: acute myeloid leukemia (AML). The authors suggest the research may point to new models in both the diagnosis and treatment of aggressive cancers, like AML. Carlo Maley Download Full Image

Maley is a researcher at Arizona State University’s Biodesign Institute and an assistant professor in ASU’s School of Life Sciences. His work focuses on applying principles of evolutionary biology and ecology to the study of cancer.

The group’s findings appear in this week’s issue of the journal Science Translational Medicine.

The cells, they are a changin’

A tumor is a laboratory for evolutionary processes in which nature experiments with an immense collection of variants. Mutations that improve a cell’s odds of survival are “selected for,” while non-adaptive cells are weeded out in the evolutionary lottery.

Genetic diversity therefore provides cancer cells with a library of possibilities, with some mutations conferring heightened resistance to attack by the body’s immune system and others helping malignant cells foil treatments like chemotherapy.

The diagnosis of cancer is often accomplished by examining a tumor sample containing many billions of cells. These are subjected to so-called next generation sequencing, a technique that sifts the vast genetic composite, ferreting out sequence variants (or alleles). The process then evaluates the frequency of these alleles, using the results to chart disease progression and assess the effectiveness of treatment.

According to Maley, such methods may obscure the true degree of genetic diversity, as well as the manner in which mutations arise.

“One issue here is that if a mutation occurs in less than 20 percent of the cells, it’s hard to detect by modern methods,” Maley said.

For example, because individual cells in the tumor probably carry unique mutations, they would be virtually impossible to observe with standard sequencing methods.

A further issue is that tracking mutations through bulk analysis of cells is typically based on certain assumptions as to how mutations arise and what their frequencies mean.

A new window

The current study attempts to provide a more accurate picture of what is taking place at the genetic level when an AML patient has a relapse or metastasis of the disease. Rather than carry out conventional bulk analysis of cells, the research group examined individual cells, screening them for the presence of two critical gene mutations common in AML, known as FLT3 and NPM1.

The results significantly alter existing assumptions of cancer progression, indicating much greater genetic diversity in AML than previously assumed. The process of convergent evolution, in which separate lineages develop similar features, appears to account for some of the observed diversity. The researchers found evidence that the exact same mutation was occurring multiple times within the same patient.

Curbing cancer’s lethality

Given AML’s near-limitless capacity for creating novel variants, what can clinicians do to halt the disease’s pitiless advance?

According to Maley, one hopeful approach would be to use cancer’s evolveability to advantage, rather than attempt to fight it head on: “Can we put pressures on the tumor that select for a behavior that we want – a manageable cancer that doesn't kill us?”

This new paradigm draws on a branch of ecology known as life history theory. The idea is to carefully study the environmental factors that may lead organisms to favor either a fast reproducing or slow reproducing strategy to maximize their survivability.

“This approach would say ‘let’s keep tumors as stable as possible and keep their resources limited.’ If we are able to keep the tumor cells contained and let them fight it out, we would expect to see more competitively fit cells that are growing very slowly,” Maley said.

While the current single-cell analysis evaluated just two mutations in AML, the results demonstrated the staggering evoleability of this form of cancer.

Eventually, researchers like Maley would like to examine whole genomes in single cells but currently many technical hurdles exist. Nevertheless, evolutionary approaches to cancer are already suggesting a broad rethinking of this complex of diseases.

A longer version of this story is available on the ASU Biodesign site.

Richard Harth

Science writer, Biodesign Institute at ASU