In the last two weeks, biomarkers have surged back into the news: Last week, a Wall Street Journal story gave notice to a Harvard study that used SELDI to identify a single potential prostate cancer marker; while this week Ciphergen announced it had data to support a possible pattern-based biomarker test for Alzheimer’s disease. Meanwhile, a study in the November issue of Proteomics described the use of protein arrays as a tool for finding biomarker patterns. ProteoMonitor caught up this week with some of the players in these recent events to talk about single marker identification versus patterns, mass spec versus protein arrays, and what is coming up for the proteomics-based diagnostics market.
Patterns Versus Smoking Guns
Brian Liu, assistant professor of surgery/urology at Brigham and Women’s Hospital of Harvard Medical School, brought proteomics into the mainstream for a day when a lead article in the Wall Street Journal featured a study in which he and other researchers identified a single potential protein biomarker for prostate cancer. The study, published in the online addition of Cancer on Nov. 3, described the use of SELDI to identify a single as-yet unidentified protein — called PCa-24 — that was present in 16 of 17 malignant tumor tissue samples but in none of the 12 control samples. “We used SELDI only because it’s a quick and easy way to generate initial profiles to look at potential changes,” Liu told ProteoMonitor.
Liu is using SELDI again to look for the marker in serum with a similarly small patient set, and plans to submit results for publication “very soon.” Eventually he will want to isolate and identify the marker, and at that point he will abandon SELDI for a MALDI MS/MS approach, he said. Liu is going straight for identifying a particular protein rather than looking for a pattern profile, he said, because “a pattern is nice, but we still don’t know what these proteins are.” Should his marker be vali-dated in a larger study, Liu plans to eventually develop an antibody-based assay for the marker that has nothing to do with mass spec. “Antibody-based assays are easier to run and you can develop diagnostic kits from a variety of methods, which one can’t do just by looking at mass spec, because the cost of the machine is horrendous,” Liu said.
But mass spec cost will not always be limiting, says Jorge Leon, who heads a molecular diagnostics consulting firm, Leomics, in Princeton, NJ. Moreover, he said, while identifying proteins is certainly desirable, finding a single ‘smoking gun’ protein that holds up to validation studies is much harder than finding a reproducible pattern. “I think most of the models that patterns capture are things that are happening frequently in cancer and are the result of multiple events that add up,” Leon said. “I would not be surprised to see a lot of those patterns [in which] inflammatory proteins, not specific enough for cancer, will probably be contrib-uting to the patterns among other molecules.” In that case, Leon said, finding a single common inflammatory marker would not be helpful. Leon also noted that patterns are more helpful for “capturing post-translational events” and that relying on patterns need not be mut-ually exclusive with identifying the proteins that make up these patterns. “I think both events are important — patterns are in my opinion an elegant way to look at the proteins that make the patterns. Then you can characterize the proteins later on,” he said.
Emanuel Petricoin, co-director of the FDA-NCI clinical proteomics program, who initially developed a pattern-based ovarian cancer test that is set to be launched by Beth-esda, Md.-based Correlogic Systems in early 2004, expressed similar skepticism about finding single protein markers. “If you can find the holy grail one, great!,” Petricoin said.”That’s great for the patients and you should do it. We don’t think it exists, but hopefully we’ll be wrong.”
Arrays versus Mass Specs: Which Will Be in a Hospital Near You?
Of course, a mass spec-based approach to biomarker discovery is not the only way to go. In this month’s issue of Proteomics, a team led by Brian Haab of the Michigan Proteome Consortium (see PM 10-31-03) published a study using arrays of proteins fractionated from prostate cancer cell lines to identify patterns of tumor-associated antigens. “These are proteins that are aberrantly expressed and mutated or presented in a different way to the immune system than is normal,” Haab said. “If we can identify that there is a group of proteins that very commonly elicits an immune response in prostate cancer patients, that immune response could be probed and used diagnostically.” Haab said that the ultimate goal would be to create a diagnostic test using the group of antibodies in the serum that reacted with the aberrant proteins.
While this approach of detecting multiple proteins associated with the immune response for use as a biomarker test is similar to the approach that Leon put forth, Leon doesn’t believe that arrays are a near-term answer. “It’s a possibility that protein arrays will become standard practice — but it will not be in the next five years,” Leon predicted. Leon said that, particularly when it came to biomarker discovery, the need to make the antibodies first rendered the arrays more trouble than they’re worth. “Mass spec can give you a profile of proteins, and it can identify the proteins …. so I don’t think we need a microarray,” he said. Leon predicted that the cost of mass specs would eventually come down to between $25,000 and $50,000 and would be made more user-friendly, so that the mass spec itself could be used in clinical tests without the need to ever make antibodies. But Leon did see a role in biomarkers for protein arrays. “Once we know that we have a panel of cytokines, for example, that are important, we need to quantify those at a very sensitive level. I think protein arrays have a better chance at quantifying something than mass spectrometry,” he said.
$3 Billion Market?
Leon predicted that the already hot biomarker market is about to get much hotter. “You’re going to see a lot of people jump up and down when this [ovarian cancer] test is announced in Q1 of next year,” he said. “You’re going to see a lot of latecomers rush in.” Leon predicted that this rush would be similar to the first rush into diagnostics and biomarker discovery among mass spec companies that came with The Lancet’s publication of Petricoin’s and Lance Liotta’s paper in February 2002. “Many proteomic companies not focused in diagnostics will suddenly become very interested in talking about a diagnostics strategy, as happened with Ciphergen two years ago,” he said. The same thing would happen this time around, he predicted, “unless the test crashes so badly that everyone thinks it’s a disgrace to be in that field”— a possibility that Leon found very unlikely.
Once similar tests for a variety of cancers are developed, Leon further speculated, the market for proteomics tests will be about $3 billion.