• The Lane Center for Computational Biology becomes SCS's newest academic unit following a summer marked by several research milestones

By Jason Togyer and Byron Spice

The newest department in the School of Computer Science has already made great strides in analyzing the biological processes that control diseases such as diabetes and asthma and identifying more than 100 genes that are potential targets for new cancer therapies.

In September, the two-year-old Ray and Stephanie Lane Center for Computational Biology became the first degree-granting program in computational biology to be included within a computer science school. It joins the Computer Science Department, Robotics Institute, Human-Computer Interaction Institute, Machine Learning Department, Language Technologies Institute and the Institute for Software Research as SCS's seventh academic unit.

Achieving departmental status is a powerful "statement of confidence" that Carnegie Mellon "recognizes and values computational biology as a discipline," says Bob Murphy, the university's Lane Professor of Computational Biology and the first head of the newly created department. "And that's different from many other places, where comp bio is included in a biology department or a medical school. Our program is different from many of those by having a higher level of computational rigor."

Also importantly, becoming an academic department allows the Lane Center to recruit permanent tenure-track faculty, Murphy says. Until now, Lane Center affiliates have had primary appointments in other departments, which has caused faculty to feel torn between doing work in their home programs and in computational biology. That tension will be alleviated, Murphy says, when the department starts appointing its own faculty, within the next year.

The center was created in 2007 with the help of a $5 million gift from former Oracle chief executive officer Ray Lane and his wife, Stephanie. At the time, Lane says, the couple felt Carnegie Mellon's computational and imaging capabilities held great promise to advance medical research. "The scientific progress we've seen since then has only underscored this belief," says Lane, chair of the university's board of trustees and a managing partner of the investment firm Kleiner, Perkins, Caufield and Byers.

The shift to department status also reflects the ongoing evolution of computational biology as a field, says Murphy, who in 1987 helped create the university's original undergraduate program in computational biology in the Mellon College of Science. (It's now a joint program between SCS and MCS, and in 1999, a master's degree was added.)

"There was optimism when the field began that if we threw a lot of computing power at biology, its problems would be solved," Murphy says. Instead, entirely new sets of problems have been created that are neither computer science nor biological science--they're specific to computational biology.

"That's exactly how interdisciplinary fields are supposed to work," says Murphy, comparing it to the evolution of biochemistry as a unique discipline, separate from either pure biology or chemistry, in the mid-20th century. "They often retain that collaborative character, but they also develop their own unique character."

The Lane Center's elevation comes after a summer in which researchers reported several milestones that could have far-reaching implications. In June, for instance, a team led by Ziv Bar-Joseph reported that gene regulatory networks in cell nuclei are similar to cloud computing networks. (See "SCS in the News," The Link, Summer 2009.) In August, another Lane Center team--led by Eric Xing, an associate professor of machine learning, language technologies and computer science--developed a statistical technique for detecting the genetic variations that contribute to complex diseases such as diabetes, asthma and cancer.

Rather than searching one at a time for genetic alterations that cause a particular symptom or trait--the conventional approach--Xing's group used a graph-guided fused lasso method to look for combinations of genetic markers, medical symptoms and environmental factors that were strongly linked to certain diseases. Severe asthma, for instance, is characterized by more than 50 clinical traits, some related to environment or activity levels, but others to symptoms such as wheezing and tightness of the chest, and still others to lung physiology.

In one test, Xing's team successfully detected a gene variant already implicated in severe asthma and then identified two additional variants that had not previously been associated with the condition.

"This approach will provide a more comprehensive genetic and molecular view of complex diseases so we can identify the genes that underlie disease processes, understand the role of genes in determining the severity of disease, and develop improved methods for diagnosing disease," says Xing, who reported the group's findings in the August 14 issue of PLoS Genetics, along with post-doctoral scientist Seyoung Kim.

The Lane Center's most important external partnership is with the University of Pittsburgh, which along with Carnegie Mellon runs a joint doctoral degree program in computational biology. (The degrees are conferred by the individual institutions.) The connection with the University of Pittsburgh Medical Center provides Ph.D. candidates with relevant expertise, avenues for data collection, and "clinically relevant" problems to work on, Murphy says.

The Lane Center is actively recruiting faculty members and discussions are also underway on the direction of the joint MCS-SCS undergraduate and master's programs in computational biology.

A major focus of current and planned research by Lane Center faculty is the development of advanced machine-learning methods for understanding how complex biological systems work. The number of variables involved in biological systems make it impossible to build models of those systems by doing experiments for every possible combination of those variables, Murphy says. That's where the machine-learning method called "active learning" becomes necessary. Active learning builds models using the currently available data, then chooses the best experiments to perform in order to optimally use those models.

According to Murphy, it represents a whole new way designing experiments: "With active learning, the design of an experiment isn't necessarily based on a hypothesis or the intuition of the experimenter, but on the model that's been created and the data that's been acquired so far."

As a result of automated learning techniques, "the Lane Center hopes to play a catalytic role in changing how biological research is done," Murphy says. "That's something we recognize is central to the university's interests, and it will have a lasting place here."