Fast, accurate cancer diagnosis? It’s gotta be in the blood.
That’s the thinking of Correlogic Systems, a Maryland bioinformatics company developing algorithms that can tease apart the differences in plasma proteins between healthy people and people suffering from cancer. With smarter software and the right set of markers, they hope, diagnosing difficult-to-catch cancers may be possible with a relatively simple and painless blood test.
Less than a year into a Cooperative Research and Development Agreement with researchers at the FDA and the NCI, the joint team has published two papers describing how their artificial intelligence technique singles out cases of ovarian cancer [see ProteoMonitor 02-18-02] and prostate cancer.
In the first study, the teams described how they used a Ciphergen SELDI-TOF system to generate protein profiles from blood taken from healthy individuals and ovarian cancer patients. Correlogic’s researchers used these profiles to identify protein abundance patterns that are sure-fire signs of cancer. Most recently, the partners were able to construct an algorithm that accurately picked out 36 of 38 test samples from men who had been conclusively diagnosed with prostate cancer. The research appeared in the Oct. 16 issue of the Journal of the National Cancer Institute.
This technique could allow for earlier diagnosis and fewer biopsies. The current standard screen for ovarian cancer, for example, only picks up about half of the easier-to-treat early-stage cancers. Similarly, the most common test for prostate cancer looks for elevated blood levels of prostate specific antigen, which must be confirmed by biopsy. Roughly 70 to 75 percent of men with moderately elevated PSA levels who undergo biopsy find that they do not have cancer.
The research team is now expanding the early test of protein patterns and taking it into clinical trials for an ovarian cancer diagnostic, said Correlogic co-founder and CEO Peter Levine — which means more work refining the algorithms. “Right now, for example, we are creating computational models using different kinds of protein chips and looking at differences between sera that was frozen, thawed, and refrozen,” he said. “There’s a huge amount of computational work that goes on — you can’t just create one model and go off to the races.”
The National Cancer Institute will conduct the clinical trials, which are scheduled to begin in April 2003. “Right now, we have a Rube Goldberg contraption that absolutely, unequivocally works,” he said. “Now the task is to make this thing work in such a way that a lab tech can do it.”
Earlier this month, Correlogic signed an exclusive licensing agreement with LabCorp and Quest Diagnostics to commercialize these tests. Levine said the company hopes to file for approval for an ovarian cancer recurrence diagnostic within one year. In the next 18 to 20 months, the Correlogic/FDA/NCI team will turn its attention to breast and pancreatic cancers, among others.
Although the research has been built on Ciphergen SELDI technology in the past, Levine said the partners are now using a higher-resolution MDS Sciex mass spectrometry machine and may switch to non-Ciphergen protein chips as well.
The method’s strength lies in its flexibility, said Levine. “We go into a research project without a preconceived notion of what we’re looking for. The question is detecting the proteins that are there, rather than saying ‘we think there’s a single biomarker, and we need to elucidate 40,000 proteins to see which corresponds to the disease state.’ It simplifies the process.”