This image depicts the acute myeloid leukemia cells treated with one of the treatment drugs. “Everyone’s leukemia is different,” said UW Medicine researcher Pamela Becker.
Dale W. Hailey/The Garvey Imaging Lab, ISCRM
Using big data to help fight a deadly cancer
Researchers built an algorithm to discern which drugs might best combat patients’ individual cases of acute myeloid leukemia.
The statistics are grim. For patients fighting an aggressive form of leukemia known as acute myeloid leukemia, or AML, doctors over the last 40 years have usually used a combination of two drugs to fight the disease, but it seldom cures it.
About two thirds of the patients initially respond to the drugs, but most relapse and the disease comes back. Only 25 percent of the patients with AML survive long term, noted Pamela Becker, a UW Medicine researcher and part of the core faculty with the Institute for Stem Cell and Regenerative Medicine (ISCRM).
“About 10 years ago, we realized that everyone’s leukemia is different from one another,” Becker said. Patient cases, however, are treated with much the same approach, in terms of first-line drug therapies.
In a paper published this week in Nature Communications, Becker and colleagues focused on key genes to fight this disease, and to do so, they needed to sort through the 17,000 genes present in each leukemia cell.
Su-In Lee, UW associate professor of computer sciences and engineering and genome sciences, and graduate student Safiye Celik created a machine-learning algorithm called MERGE (Mutation, Expression hubs, known Regulators, Genomic CNV and Methylation) to test biomarkers and their response to 160 cancer drugs. The algorithm predicts which drugs are the likeliest candidates to successfully treat each AML patient.
With the algorithm, Lee was able to identify a few dozen genes that might predict sensitivity or resistance to certain drug classes.
“The question we were trying to answer is whether we could use level of expression of certain genes to reveal vulnerability to certain drugs and enable new treatment for patients,” Becker said.
“Drug development is an expensive and challenging process, and cancers that appear pathologically similar can respond to the same drug regimen in different ways,” said Lee. “There are more than 1,200 potential cancer medicines in development in the United States alone. We need better methods for matching patients with the most effective treatment, and this has been our goal with MERGE.”
Lee and Becker determined that about 100 of the 17,000 genes were more meaningful, and that about a dozen of those would give them answers to which drugs – existing or new – would work best with individual patients. By correlating which genes responded to which drug, Becker’s research opens up new possibilities for treatment, perhaps in drugs not considered before.
Becker says she plans to start a clinical trial soon using this new data on patients.
Other researchers included C. Anthony Blau, UW professor of medicine (hematology); Vivian Oehler, UW associate professor of medicine (hematology) and Fred Hutch researcher; Tim Martins and James Annis from the Quellos High Throughput Screen Core Laboratory at ISCRM. Sylvia Chien and Jin Dai from the Becker lab worked on all the samples for this project.
The work was supported by the National Institutes of Health (T32 HL 007312), National Science Foundation, American Cancer Society, Life Sciences Discovery Fund, National Cancer Institute of the National Institutes of Health (P01CA077852), and philanthropic funding from Norman Metcalfe.
source: University of Washington