Carcinoma of the ovaries is a rare malignancy faster acting and deadlier than breast cancer. These paired, reproductive, egg-laying glands don’t tip their neoplastic hand until their tumors are well entrenched, and beyond effective treatment.
By the time vague symptoms of dyspepsia, bloating, early satiety anorexia, gas pains and backache hint at ovarian cancer, more than 80 percent of the patients will have advanced cases when diagnosed. Only one in four of them will live five years or longer. But if detected early, 90 percent can beat the five-year survival limit, and most are cured by prompt surgery. Last year, 23,400 women in the U.S. developed ovarian cancer (OC), and 13,600 died of the disease.
“A number of different risk factors are associated with OC,” observed molecular biologist Emanuel Petricoin III, at the Food and Drug Administration’s Center for Biologics. “They’re linked with genetic predisposition familial issues. Like breast cancer, most ovarian cancers are considered sporadic, independent of any inherited predispositions. In some ways,” Petricoin added, “OC is an orphan disease, because it’s 10 times less frequent than breast cancer. But it’s critically in need of early intervention and a single good diagnostic biomarker that can make a big difference in terms of therapy.”
Petricoin is first author of a paper in the current Lancet, dated Feb. 16, 2002, titled: “Use of proteomic patterns in serum to identify ovarian cancer.” Its senior author is Lance Liotta, chief of pathology at the National Cancer Institute. He and Petricoin are co-directors of the NCI/FDA Clinical Proteomics Initiative.
“Imagine the pathologic condition taking place inside of an internal bodily organ in this case the ovary,” Liotta told BioWorld Today. “The blood is perfusing to the ovaries in the pelvis, so we thought there might be modifications in the proteins in the blood. We could then look at the peripheral blood, take a sample, and find patterns that reflect the pathology of the underlying organ.
“That hypothesis,” Liotta went on, “was supported by our paper in the Lancet, and it has great ramifications for proteomic patterns in blood that might reflect any potential disease state. Our goal is to verify this and extend it to other diseases. The general advantage of our computer learning system is that as we enter more data into it over time, the proteomic pattern keeps getting bigger, better and smarter.”
One Drop Of Blood Parses Millions Of Proteins
This protein pattern-recognition diagnostic begins with a single drop of blood taken from a patient’s finger by needle-stick.
“We use a genetic algorithm based on bioinformatics methods,” Petricoin explained. “It’s like an artificial intelligence-based computer program, which searches hundreds of millions of combinations of protein patterns by mass spectral data, until it finds in a first-phase training set’ the protein pattern that best distinguishes normal healthy patients’ sera from patients with ovarian cancer.”
Petricoin continued: “We first trained’ the algorithm on what’s already known. That is, sera taken from healthy women followed for five years after serum was collected vs. women whose serum was collected when they had active ovarian cancer. We trained the algorithm on those serum samples, and found hundreds of millions of iterations the optimal pattern fit. It searched and searched until it found a protein pattern that distinguished the ovarian cancer from the healthy. It did this using mass spectra data streams.”
“The system we used to generate the proteomic spectra,” Liotta pointed out, “is a type of mass spectroscopy called Surface-Enhanced Laser Desorption and Ionization. “The way it worked,” he recounted, “is we had a metal bar coated with chemical that binds a subset of proteins in the blood. We put our drop of blood on that surface and some subfraction of the thousands of proteins in the sample stuck to it. Then we put that bar into a vacuum tube, and shot it with a laser beam that blasted the proteins off the surface. They slid down a vacuum tube and stuck to an oppositely charged electrode. The small proteins slid fast, and the big ones more sluggishly. Looking at the blips coming off that electrode gave us a measure of the proteins’ relative mass and size charge.
“It’s a spectrum of all the proteins that are flying off the surface of the tip,” Liotta went on. “So that comes out like a bar code of proteins. They’re predictive, so when we tested an unknown patient sample, only those particular protein areas were inspected to see if the patterns showed likely cancer, likely noncancer or something new.”
The co-authors first analyzed serum samples from known OC patients and unaffected individuals, to establish proteomic patterns present at different levels in the two cohorts. Then they compared them with the same protein patterns in sera from other patients with and without cancer. The team correctly identified 50 out of 50 cancerous samples, and 63 of 66 noncancer samples.
Proteomic Patterns Target Bottom Line
“Almost all currently licensed FDA therapeutics target the proteins, and modulate protein activity,” Petricoin pointed out. “It’s very important for both the NCI and the FDA to use proteomics in these applications. Proteins for the next decade,” he foresees, “are going to be the majority of the drug targets, imaging targets and biomarkers. Proteomics help because they’re very proximal to the bottom line.
“We’re employing this technology,” Petricoin observed, “not just for ovarian cancer detection, but actively pursuing intriguing applications in other types of cancer breast, prostate, colon, lung, pancreas. And our program is looking at detection of infections, such as anthrax or other pathologic agents all from serum.
“Right now we have a clinical trial at the NCI,” Liotta said, ”that’s open to accrual. We’re enrolling women treated for ovarian cancer, and evaluating them for the potential of relapse. So we’re looking for proteomic markers in serum that can help detect relapse earlier. We’re also going to be opening new clinical trials for high-risk screening as well in the next few months.”
Petricoin noted, “The U.S. government has patented this proteomic concept along with Correlogic Systems Inc. in Bethesda, Md., which developed some of our underlying algorithms.” To which Liotta concluded, “It’s just possible for some biotech entity to license these patents from Correlogic and the government.”