Skip to main content

Diagnosed With Ovarian Cancer, a Researcher Mined TCGA Data to Study Her Own Disease

, by Amy E. Blum, M.A.

Shirley Pepke, Ph.D., with her collaborator, Greg Ver Steeg, Ph.D.

Credit: Gus Ruelas/USC News

A computational biologist, wife, and mother of two, Shirley Pepke, Ph.D., studied regulatory genomics until she was abruptly diagnosed with stage IIIC ovarian cancer. In a struggle against time, Pepke started to focus her scientific expertise on her own cancer.

Researching for Her Life

Given the bleak prognosis associated with her stage of ovarian cancer, Pepke realized that traditional therapies were not likely to provide her with much more time with her husband and kids. By applying her knowledge of genomics, Pepke thought that she might be able to tailor her treatment to her particular cancer.

Pepke had experience with big genomic data; she had participated in the ENCODE project. But she had never directly worked on cancer research, and she would be undergoing intense chemotherapy.

“It didn’t seem likely that I would make any headway given the circumstances,” explained Pepke, “but I had to try.”

Pepke’s close friend from graduate school introduced her to Greg Ver Steeg, an assistant professor at University of Southern California who specializes in analyzing big data. Ver Steeg offered to help Pepke study ovarian cancer using a machine-learning algorithm he developed called CorEx.

“I’m really grateful to Greg and impressed by his bravery,” Pepke said. “Perhaps all researchers want to have an impact, but it’s different to work with someone who might use your research as part of their medical decision."

While Pepke was receiving front-line chemotherapy, she and Greg began mining data on ovarian cancer and looking for patterns. Because ovarian cancer is not defined by a large number of DNA mutations, they focused on gene expression by analyzing TCGA data about RNA.

“I came to realize that the TCGA dataset was particularly valuable because it was the only large-scale RNAseq dataset,” said Pepke. “There were lots of micro-array data available. But when I compared RNAseq data to the microarray data, looking at the RNAseq data was suddenly like de-fogging a glass. There was so much more depth and detail of signal that it was really striking.”

Pepke and Ver Steeg used CorEx to look at genes in TCGA that were co-regulated in ovarian cancer, meaning that their expression levels in the cell appeared to be linked. Then they zeroed in on sets of co-regulated genes that correlated with better or worse survival.

“When I did the survival analysis certain things came out very clearly,” said Pepke. “CorEx parsed the expression of genes involved in the immune response into subgroups and particular subgroups correlated with long-term survival following chemotherapy.”

Confronting Her Own Cancer Data

Unfortunately, Pepke had an unusually short remission after chemotherapy.

“I had recurred so quickly, so I wanted answers,” she said. “I started looking through my own data because I felt like there had to be an answer there and it was up to me to find it.”

It was then that Pepke noticed something about her cancer. When she compared the gene expression of her tumor to other ovarian tumors in the TCGA dataset, she saw that her tumor had unusually high expression of genes involved in the immune system, which usually correlates to a sustained response to chemotherapy. But in her case, it didn’t.

“I knew that I had an immunogenic tumor, but my immune response wasn’t being effective enough to help me,” she explained.

Choosing Immunotherapy

Combining what she learned from her analysis with the fact that chemotherapy was not effective against her cancer, she decided that a checkpoint inhibitor immunotherapy, which may boost her immune system’s ability to attack her cancer, might be worth trying. Pepke consulted with immuno-oncologists who reassured her that the side effects were likely to be manageable and that the treatment could be effective. Pepke’s oncologist also found encouraging pre-clinical studies in which immune checkpoint inhibitors helped to effectively treat human cancers that were transplanted into mice.

With her life on the line, Pepke carefully weighed the potential for sustained success of the different treatment options with their side effects. She opted to take the checkpoint inhibitor, have her tumor removed surgically, and undergo chemotherapy. But after all this, her cancer progressed. Without promising options for further treatment, Pepke decided to stop her chemotherapy to enjoy the summer with her family and let her body recover. She returned to the oncologist in the fall expecting the worst.

“Basically my husband and I were thinking, how bad is it going to be?” said Pepke. “But my oncologist came in and said ‘well something’s working, because your tumor marker has fallen to less than half of what it was.’”

Her tumor marker continued to decline into the normal range and imaging revealed no cancer. That was a year ago, and Pepke is still cancer-free.

Ever the scientist, Pepke is realistic about her exceptional response.

“This type of response is so rare my oncologist almost didn’t believe it, “she said. “We also don’t have enough information to know if the checkpoint inhibitor made a difference, but the time course was consistent with an immune response.”

With New Life, a New Mission

Grateful for the people who helped her and the valuable time she has gained, Pepke continues to devote herself to translating ovarian cancer genomic data to the clinic, where she aims to benefit other patients like her.

“There is so much that we don’t understand about cancer that we need to push things further and not be discouraged in the face of the complexity,” she urged. 

She also wants to see more women receive the kind of personalized care that she benefitted from as an exceptionally well-informed and connected patient.

“Just because something is not informative for most patients, does not mean it won’t be for an individual patient,” she explained. “I think we should give personalized care. Combining patient-specific features with database information can make a difference. With an individualized approach, more women might experience the same happy surprises.”

Featured Posts