SAN FRANCISCO — What will it take to move proteomics-based diagnostic biomarkers out of the lab and into the physician's office? According to researchers here this week at Cambridge Healthtech Institute's Beyond Genome conference, performance and usability will be the ultimate drivers for clinical acceptance.
"It is not the FDA's approval that's going to make the clinician use [a biomarker], or that the study was published in a famous journal, or a company that is going to convince them to use it," said Daniel Chan, a professor and the director of the Biomarker Discovery Center at Johns Hopkins University. "The people developing the biomarkers need to address what the clinical needs are, and how well the biomarkers perform in terms of sensitivity, specificity, and ROC curves."
Chan spoke during a panel discussion on proteomics in clinical research and diagnostics at the conference.
Last August, Chan published a paper in Clinical Chemistry describing how his research group used Ciphergen's SELDI technology to select a profile of biomarkers for ovarian cancer, then narrowed the biomarkers down to the three most promising candidates: a truncated form of transthyretin, apolipoprotein A1, and a cleavage fragment of alpha-trypsin (see ProteoMonitor 8/20/2004).
Chan's group has also discovered two biomarkers for prostate cancer that, when combined with the standard PSA prostate cancer biomarker, significantly improve sensitivity and specificity, compared with the prostate cancer diagnostic test based on PSA alone.
Chan noted that though both the ovarian cancer and prostate cancer biomarkers seem promising, clinicians are not likely to adopt them if they have no personal experience with them, even if the markers are approved by the Food and Drug Administration.
"It's important to get into the clinic early as a team," said Chan. "If clinicians have already tested it as part of clinical trials, as soon as it's FDA approved, they will start using it. If they've just read about it in a journal, they're not likely to use it."
Brad Guild, the senior director of protein sciences and structural biology at Millennium Pharmaceuticals, agreed with Chan.
"The sooner you start a dialogue with a clinician, the sooner you're going to be on the same page," he said.
Clinicians are looking for specific markers to answer specific questions, Chan added. A predictive marker that can tell whether a patient will respond to a certain drug is more useful than a non-specific early diagnosis screening test that opens up more questions.
Randall Nelson, the president and CEO of Intrinsic Bioprobes, noted that although diseases affect populations differently, that difference is rarely taken into account when designing drugs and diagnostics.
"Diagnostics are not segregated into age, race, sex. That's highly overlooked," he said. "It's one-size-fits-all when it comes to diagnostics."
Nelson and his research team at Intrinisic Bioprobes study "population proteomics," which takes into account the proteomic variation between individuals in a population. In one pilot population proteomics project, the team analyzed 25 proteins from 96 samples representative of the general population of the United States using 25 antibodies. Their goal was to determine the number of structural variants and modifications of the proteins that exist within the general population in order to establish a basal level of variations in the proteome that exist in the general population.
Now the Intrinsic Bioprobes team is expanding the population study to 5,000 individuals.
"If we can look at 5,000 individuals, then we can at least get an idea of the boundaries and error bars on one individual," said Nelson.
Chan noted that in designing biomarkers, it is important to minimize and control for both inter- and intra-individual variability. One reason Chan's research team does not use mass spectra pattern profiles by themselves as biomarkers is that the spectra vary too much with different time points.
"If you identify the proteins that are part of the panel of biomarkers, then you can at least study the proteins and their variation," said Chan. "We have to know more about [the protein biomarkers] and design diagnostic panels that don't vary. If they do vary, the important thing is to know it and to control it."
Lorah Perlee, the director of scientific applications at Protedyne, spoke about steps that her company took to standardize a system for performing multiplex protein assays in order to discover biomarkers.
"Biomarker discovery is not achieved in a single path," Perlee said. "We had to do multiple iterations with multiple sample sets before we could guide decision making."
Some questions that were addressed along the biomarker discovery path had to do with experimental design, such as whether there were enough patients to sufficiently test the hypothesis, and what to do if there were sample shortfalls.
One of the most important steps in standardizing biomarker discovery was bringing together three main functional units: client services, manufacturing, and service procedures, Perlee said.
The company had to determine project hand-off points, procedures for dealing with the failure of reagents, and procedures for dealing with how samples are shipped and how clinical information is entered into the company's laboratory information management system.
Ultimately, with a standardized system, Protedyne was able to increase its project load from 25-50 projects to 200-500 projects, Perlee said.
— Tien-Shun Lee ([email protected])