Doctors have nearly a dozen new targeted drugs to treat patients with acute myeloid leukemia, or AML, yet three of four patients still die within five years. Some patients succumb within just a month or two, despite the battery of drugs used to treat the aggressive blood disease, where blood cells don't develop properly.

A new study draws on a field of science known as proteogenomics to try to improve the outlook. In a paper published Jan. 16 in Cell Reports Medicine, scientists report new findings about how drug resistance in some AML patients develops and how doctors might someday stop or slow the process.

The research comes from a team of researchers from the Department of Energy's Pacific Northwest National Laboratory and Oregon Health & Science University. For nearly a decade, OHSU and PNNL researchers have worked together to fill a critical gap in our knowledge of how cancer and other diseases happen. At one end of the spectrum, our body's genes can go awry, creating mutations that can be harmful or deadly. At the other end of the spectrum is a real person whose life is affected or even ended as a result.

What happens in the middle, between the genes and the person's health?

The answer: a dizzying number of complex molecular processes that scientists are grappling to understand. At the center are the body's proteins and a field of study known as proteogenomics.

Sorting the data with machine learning

The PNNL-OHSU team is studying thousands of proteins that could play a role in AML. Proteins are the body's molecular workhorses, ferrying nutrients and other supplies back and forth between cells, turning genes on or off, and maintaining dozens of basic body processes.

Even though genes get the glory, they do little directly to keep our bodies going. That's the job of proteins. For nearly 20 years, study author Karin Rodland of OHSU, formerly of PNNL, has been a pioneer exploring the role of proteins in health and disease, building a program with OHSU and PNNL colleagues to study AML.

In the latest study, a team led by Sara Gosline, a data scientist and computational biologist at PNNL, did an exhaustive study of the protein activity in 210 patients with AML. Altogether, the team measured levels of nearly half a million pieces of proteins from more than 9,000 proteins in patients' blood samples.

The team combined those findings with extensive data already known about the disease—the genes and mutations involved, the molecular messengers that indicate which genes are active, and the effects of 46 drugs on AML patients, along with information about how the disease progressed in those patients.

"We were able to look at patterns of drug responses in hundreds of people by including protein and gene measurements together, and this gave us a level of detail that hasn't been possible in prior studies," said Gosline. "This is a great example where we are able to put our growing knowledge of protein signaling and machine learning models to benefit patients in the future."

Gosline and colleagues, including first author James Pino of PNNL, deployed artificial intelligence, using several machine learning algorithms to make sense of the data.