by Jeff Mortimer

Nana Lee is a researcher with a mission. Like a gunslinger in an old Western movie who’s tracking the man who shot his pa, Lee is after the disease that claimed her mother’s life in 1997.

“It’s not just science but also a personal project,” says Lee, a post-doctoral fellow in internal medicine who is part of the colon cancer profiling research team in Eric Fearon’s laboratory. “It’s something that drives me with my work.”

“They took out her whole colon but they didn’t know, they couldn’t tell me, if it might have metastasized,” she recalls. “It was just a big question mark. They said it might have spread to the lymph nodes, but they weren’t sure; it was tricky to diagnose. Eleven months later, they said it had spread to her liver. She died a couple of months after that. I knew that if we had had the tools to diagnose if it had metastasized sooner, things would have been a lot different. Now one of my goals in life is to come up with something like that.”

As Lee was finishing her Ph.D. in biochemistry at the University of Toronto, she learned about the colon molecular profiling project at Michigan, headed by Eric Fearon, M.D., Ph.D., and Stephen Gruber, M.D, Ph.D., “and I just jumped on that. I came here last September because they had a great program and I wanted to study genomics and proteomics of colon cancer and help find a way to use this information to improve clinical practice.”

While not as personal, similar passions inflame other researchers working on the cancer molecular profiling project. Kathleen R. Cho, M.D., the principal investigator of the ovarian profiling cancer project and an associate professor of pathology and internal medicine, is a practicing surgical pathologist with a subspecialty expertise in gynecological cancer diagnosis. “I spend much of my life dealing with these kinds of tumors,” she says. “It’s exciting for me to get beyond the microscopic appearance of these tumors and begin to evaluate their molecular profiles. I’m optimistic it’s going to take us way beyond what we’ve been able to do with just microscopic appearance alone.”

For Cho, too, early diagnosis is the holy grail. “It’s been a huge problem [with ovarian cancer],” she says. “Tumors tend to present very late in the clinical course, with high-stage disease. If they present with low stage disease, they can often be cured with surgery alone or surgery and chemotherapy, but if they present with high stage, they’re very difficult to cure with the treatment modalities we have currently.

“The job now is to ferret out, among the hundreds of proteins and thousands of genes we’re looking at, which of those individual molecules may be the most predictive in determining a particular tumor’s biological behavior,” says Cho. “If we can identify genes that are very highly expressed specifically in ovarian cancer that might be, for example, secreted into the blood, we could develop a simple blood test to enhance early diagnosis.

That would be the long-term goal, to develop screening tools that would allow you to identify patients with low-stage ovarian cancer by doing some sort of simple, minimally invasive test, like a pap smear for uterine cancer, that’s convenient and not frightening or costly.”

Good Prognosis, Bad Prognosis: Genes and Proteins Tell the Story

The lung cancer profiling team is headed by David Beer, Ph.D., associate professor of surgery and radiation oncology.

“We’re trying to determine whether we can identify the genes and the proteins which are associated with a poor clinical outcome in early-stage lung cancer,” says Beer. “Most of the patients who have stage-one lung cancer will do well, but about 25 percent will have a poor clinical outcome.”

So far, so good, says Beer. “We’ve been able to identify a large number of genes and also some specific proteins which seem to distinguish tumors which have a bad prognosis from ones that have a good prognosis,” he says. “The proteomics studies are aimed at trying to distinguish not only patient prognosis but also the types of proteins which are potentially unique or highly expressed in tumors with different clinical features, such as invasive characteristics, and also the proteins that are associated with the ability to metastasize to local lymph nodes.”

For example, he says, “You could look at two different tumors under the microscope and you can’t tell the difference between them, but the genes that are expressed and the proteins that are encoded by those genes are expressed differently. Careful quantitation of the levels of those proteins, as well as their identification, gives us a tool to try to distinguish tumors in a way that you just can’t determine by looking at the tumor histologically.”

The day when such tools are clinically available may not be far off. “It will probably start showing up fairly soon,” Beer says. “We’ve identified those genes and proteins which are of interest and may be useful. The next step will be to bring this to the clinic to test it in a prospective manner by taking many more earlystage tumor patients and repeating this study that we’ve just done with nearly a hundred and see if the markers that we’ve identified truly do define their prognosis.”

A New Way of Doing the Numbers

The “careful quantitation” to which Beer refers is founded on the work of a team of statisticians and image analysts headed by Jeremy Taylor, Ph.D., professor of biostatistics in the U-M School of Public Health, and including Sharon Kardia, Ph.D., assistant professor of epidemiology there; George Michailidis, Ph.D., assistant professor of statistics; Kerby Shedden, assistant professor of statistics; and Rork Kuick, M.A., a statistician who has worked on genomic and proteomic quantitative analysis for a number of years.

“We barely had machines to do this work back then, of course,” says Kuick. “It was only at the very end of the 1980s that computer windowing systems permitted spot detection and quantification algorithms to allow a user to work with many, many images and get this work done. To make even marker discovery work, there’s a lot of image analysis involved. The images have to be obtained and spots measured with algorithms and spots between different patterns matched with algorithms, so just obtaining the data is a bit of an engineering and computer science task.”

The field of statistics, as well as medicine and science, has benefited from the challenges involved. “Some of these new technologies require new methodologies in order to look at the data and extract the most out of them,” says Michailidis. “What you would like to do is make these comparisons and see which genes have changed significantly. One of the important statistical calculations is to quantify precisely what you mean by significant difference. Say a normal tumor sample is 1 and we find one whose expression level is 1.1. Is that a significant expression? If it went from 1 to 15, you’d say this is really large, but from 1 to 1.1, is it such a big difference or not? That’s where statistics come in, to see if differential expression is really there.”

Moreover, he adds, “The data are fairly noisy. It’s a complicated technology and there are lots of sources of errors, so you need to correct for as many of these errors as possible to see if you really have a significant difference or not. If you have, then you go to the next stage and try to learn more things about these changes.”

Project statisticians are also working with a relatively small number of samples, due to the cost and difficulty of obtaining them. Michailidis says this “leads to a totally new paradigm in statistics. Usually what we have are fairly large samples and a few variables; this is the other way around from what we are used to. Instead of 1,000 samples and five variables, we have 100 samples and 7,000 variables. This changes the game in fundamental ways. Old methodologies don’t apply. It becomes fairly exciting because there is a lot of room for new ideas.”

The entire biomarker enterprise is energized by connections and collaborations, not only among medical specialties but also between medicine and other disciplines. The project’s individual actors are intrigued by the whole that their parts comprise. “To be honest, I don’t understand the underlying science that well,” says Michailidis, “but I’m brushing up on my biology.”

Says Cho, “I’m absolutely turned on by this. It’s been a great opportunity to collaborate with other investigators doing similar work with other tumor systems.” The mission that unites them is finding out more. As Beer says, “The more we learn, the more new targets we can potentially uncover for both diagnosis and therapy.”

Also:

Beyond the Genome