Rice, MD Anderson lead leukemia crowd-source project

Institutions administer DREAM challenge to accelerate search for treatment 

A Rice University bioengineer is leading an international competition to improve the analysis of genetics and proteomics to help leukemia patients.

Amina Qutub, an assistant professor of bioengineering at Rice’s BioScience Research Collaborative (BRC), is scientific lead on this year’s ninth annual DREAM challenges, an online, crowd-source challenge to the systems biology community.

The BioWheel invented at Rice as a way to visualize massive amounts of data will be part of this year's DREAM challenge to improve the analysis of data related to leukemia. Courtesy of the Qutub Lab

The competition will kick off with a launch party at the BRC June 2 at noon. The event is sponsored by the Houston Area Translational Research Center and the Texas Medical Center (TMC). Strengthening relationships with TMC institutions is one of Rice’s Priorities for the New Century.

Rice and researchers at the University of Texas MD Anderson Cancer Center (MDA) are organizing the challenge in collaboration with the nonprofit Sage BioNetworks. During the open-challenge phase of DREAM (which stands for Dialogue on Reverse Engineering Assessment and Methods) from June to mid-September, researchers will study the same sets of data and mine them to create the best analytical tools to draw conclusions.

In one of this year’s DREAM challenges, an expected 300 participants will look at extensive data gathered over eight years by Steven Kornblau, a professor in the MDA Department of Leukemia, from acute myeloid leukemia patients.

In addition to traditional diagnostic, clinical and treatment information, the data set generated with a grant from the Leukemia and Lymphoma Society incorporates protein expression and activation state information for patients treated by Kornblau and his MDA colleagues. That makes it a unique resource, he said.

“It’s one of the most extensive data sets in the world for leukemia,” Qutub said. “Steve measured more than 300 different variables for each patient.” The contestants’ task will be to crunch the raw data and predict whether a patient would likely recover or relapse, as well as their overall length of survival.

Qutub suggested the challenge last year at a conference where she presented her lab’s winning 2013 project, the BioWheel, a graphic tool to visualize protein networks and treatments for breast cancer. “I was talking with the DREAM chair, Gustavo Stolovitzky, about the data set for last year’s challenge, which was from Gordon Mills’ group at MDA on cancer cell lines,” she said.

“I mentioned to him, ‘We’re sitting on a set of actual patient data with a collaborator at MDA,’ and suggested we could give people access through DREAM and find models to predict which drugs or which pathways to target to help patients.”

Kornblau welcomes any opportunity to “see the larger structure to the data that I always thought was there. It’s been a learning process to get there and figure out how to see it.

“Our patients and our physicians are pretty hungry for innovation,” he said. “Things might move quicker than we anticipate if we get good results and there’s actionable information,” he said.

Members of the DREAM-winning BioWheel team in Rice's Qutub Lab, from left: Wendy Hu, Byron Long, Amina Qutub, Alex Bisberg and David Noren.

“Steve Kornblau is a hero for agreeing to make the best use of this data through a challenge,” said Stolovitzky, founder of the DREAM project and the director of the Translational Systems Biology and Nanobiotechnology program at IBM’s Thomas J. Watson Research Center. “I understand there is a lot of effort in creating these data sets, and very often the data can sit in silos where only one or two researchers have access to it. But we feel data sets produced with federal funding, especially when they are related to patients, belong to everybody.”

Stolovitzky, who is also affiliated with the Department of Biomedical Informatics at Columbia University, sees the DREAM project as serving four purposes: To find the best algorithm for a particular problem, to encourage collaboration between the competitors, to democratize the data and to accelerate research that improves the lives of patients.

“We let everybody who cares analyze that data and try to solve important biomedical problems,” he said. “In that sense, we are making the data multiply its impact. By getting many teams working in parallel, they can do in four months what might take a single team years.”

Because Qutub’s lab is working with DREAM researchers to develop the scoring system by which the competition will be judged, the Rice team won’t go for two wins in a row. “The challenge is to focus on predicting relapse or remission durations, so we’ve been testing the data to figure out current best predictions, the test set and a scoring mechanism. Basically, we have the answers,” she said.

However, the Rice BioWheel will be present. “Last year’s data was given out as an Excel chart, but this time we’re visualizing the data set,” she said. “Teams can interact on the website using BioWheel to see the data, touch the data, play with it. They can put their predictions up and see how they match survival curves.”

The organizers are negotiating to publish their results in a major journal soon after completion of the competition. “The writing starts when the challenge starts,” Qutub said.

Sponsors of the competition include Rice, MDA, the IBM Computational Biology Center, Data Is Beautiful Solutions and Sage BioNetworks.