SAN DIEGO — Following a talk by Leigh Anderson that showed that the number of new protein-based diagnostic tests that have been approved by the US Food and Drug Administration has declined over the last 10 or so years, scientists here at the Cambridge Healthtech Institute's PEPTALK conference discussed why proteomics has not yet produced a biomarker that has made a significant clinical impact.
Anderson, the founder and CEO of Washington DC-based Plasma Proteome Institute, showed that since 1994, the average number of new protein-based tests introduced per year was close to three. Since 2004, however, the average number of such tests has been less than one.
Some scientists said cost is an issue when it comes to developing a successful, FDA-approved protein biomarker-based test.
Anderson estimated that the cost for developing an immunoassay is $2 million to $4 million per protein target.
Joanna Hunter, the associate director of protein analysis at Caprion Pharmaceuticals, who moderated the discussion on biomarkers, said the issue was a bit of a "chicken and egg" problem.
"Government foundations want an example where proteomics has cured a patient," she said. "They say, 'Show me one example and I'll put billions of dollars into proteomics.' Well, which comes first, the proteomic-based cure, or the billions of dollars?"
"Government foundations want an example where proteomics has cured a patient. They say, 'Show me one example and I'll put billions of dollars into proteomics.' Well, which comes first, the proteomic-based cure, or the billions of dollars?"
Other scientists cited scientific challenges as the reason that proteomics hasn't been more successful in this realm.
"Disease heterogeneity is absolutely tremendous," said Samir Hanash, director of the molecular diagnostics program at the Fred Hutchinson Cancer Research Center. He pointed out the fact that lung cancer refers to many types of the cancer, including small cell, non-small cell, smoking-related, and non-smoking related. "Maybe we're asking too much for a biomarker to serve for all these types of lung cancer."
Hanash added there is also a lot of heterogeneity among people, which can lead to "confounding factors" in control patients.
"What is 'normal?'" asked Hanash. "The intravariation among the control group is often higher than among the disease group."
Christopher Becker, executive director of Wilmington, NC-based PPD, noted that retrospective studies based on samples that have been sitting in the freezer may not be successful because the collection procedure for the samples is not known, or was not well standardized. At the same time, prospective studies are expensive and take more time.
Becker added that it takes time for biomarkers to be developed into successful tests.
"Validation doesn't happen overnight," said Becker. "Proteomics is a pretty new field. There's a lot of heat, light, and noise going on, and it takes some time for the cream to rise to the top."
Other scientists hypothesized that it may be harder to develop successful tests these days because whatever tests are developed must perform better than tests that are already in existence.
Prior to the discussion on the lack of proteomics-based biomarker success, John Yates III, a professor of cell biology at the Scripps Research Institute, highlighted some challenges in dealing with the plasma proteome.
In his talk, Yates shared some terms that had been coined by Harvey Pollard, a professor at the Uniformed Services University of the Health Sciences, to describe various proteomic strategies, and their shortfalls.
"There's the Casablanca approach, where you round up the usual suspects," Yates said. "Then there's the Lamp Post approach, where you look under the light because you can see there."
Other proteomic approaches included the Transportation Security Administration approach, where "you search everyone;" the Magical Induction approach, where "if enough data is collected, the answer will appear;" and the Low IQ Proteomics approach, where "you do easy experiments and depend upon magical induction."
On a more serious note, Yates pointed out that the dynamic range of protein concentration within serum or plasma is around 1011, and the dynamic range of protein concentration in tissues is around 108.
The enormous range in protein concentrations leads to the challenge of enriching for proteins that are likely to serve as good biomarkers. Those proteins are generally believed to exist in low concentrations within biological fluids.
Yates also noted that results from proteomic experiments are often not reproduced from lab to lab, or even within labs. This could be because of a difference in instruments, software, and/or data analysis, Yates proposed. Or it might be due to biological variability between samples.
On a positive note, Yates pointed out that improvements in proteomic technologies have led to improved success in finding promising disease biomarkers.
"Technology development is still critical for proteomics," he said. "I worry that with plasma, the technology is not quite there yet, and that may discourage funders."
Anderson, who wrote a paper analyzing the diagnostic prospects of the human plasma proteome, published in Molecular & Cellular Proteomics in 2002, said that the FDA had one thing to say for certain when asked why the number of new protein-based diagnostic tests had declined to less than one per year.
"The one answer that they could give with absolute assurance was that it was not their fault," said Anderson.
— Tien-Shun Lee ([email protected])