At a roundtable discussion on biomarkers held during last week’s Association of Biomolecular Resource Facilities conference, researchers gave some practical tips for handling samples, pointed out some special challenges in dealing with the plasma proteome, and discussed the pros and cons of using spectral patterns as biomarkers.
When collecting samples from patients, researchers should be careful to keep a consistent protocol and to minimize biological variation due to activities such as eating, Sushmita Mimi Roy, a researcher at Menlo Park, Calif.-based Surromed, said during the ABRF discussion in Savannah, Ga. For serum samples, this means collecting blood after patients have fasted. For spinal fluid, Roy suggested that the first draw be discarded in order to avoid serum contaminant.
Roy also said that all samples should be stored at -80°C until they are ready to be processed. The number of aliquots needed for experiments should be determined early in the process so that they can be separated out and kept free of contamination.
“This is just some real practical advice for biomarker discovery,” said Roy.
To measure how well samples keep at -80°C, Roy analyzed 400 different proteins quantitatively, then requantified the same 400 proteins after they had been stored at -80°C for one year. She found that of the 400 proteins, only 44 had degraded significantly over the year, resulting in greater than one fold decrease in concentration.
Echoing what Kevin Coombes, a biostatistician at MD Anderson Cancer Center (see Proteomics Pioneer, p. 6), said during an earlier talk, Roy cautioned that the run order of samples should be randomized in order to minimize skewing of results due to instrument variability. In randomizing samples, researchers should consider pairing matched samples, such as control and disease samples.
“You have to maintain a system and to run quality control checks,” said Roy.
Turning to the topic of plasma proteome challenges, Gregory Opiteck, a researcher at Caprion Pharmaceuticals, pointed out that the dynamic range of a mass spectrometer is only about one part per thousand, while the dynamic range of the plasma proteome is about 1011. This results in difficulty in detecting low-abundance proteins because of a large signal-to-noise ratio.
“The plasma proteome is very noisy,” said Opiteck. “It may matter more what your suspect had for lunch [than what disease he or she has].”
Opiteck suggested using mice as an alternative to humans because they have diseases that are easier to track and have more controlled variability.
“The advantages of mice are that you can make uniform species a priority, you can backtrack to earlier stages of the disease, and you can use patients as their own controls,” said Opiteck.
In trying to determine how many patient samples are needed to find a biomarker, Opiteck suggested using a rule of thumb that has been proposed by statisticians: Collect at least 30 samples per disease state or dose.
Another tip that Opiteck gave was to make sure that enough sample is collected so that the biomarker can be detected.
“If you’re looking for a protein like PSA, for example, which is present at about one nanogram per milliliter, you have to make sure you have enough plasma [to see the biomarker] — in that case about 30 microliters,” said Opiteck.
He suggested that depleting the most abundant proteins in serum in order to better see lower-abundance proteins is a good idea, even if there is a risk that some proteins may be depleted with the high-abundance proteins.
“You’re already assuming that biomarkers exist, that they’re proteins, that they’re in the protein database, that they’re in the literature,” said Opiteck. “Why not try depletion of albumin? That the protein doesn’t bind to albumin is just another condition that you’re assuming exists.”
Opiteck then addressed another question posed by the biomarker roundtable — whether patterns of spectra are useful as biomarkers. He suggested that patterns should be used to guide biomarker discovery, but that patterns without identification are not sufficient to serve as biomarkers.
“It’s logical to collect spectra and to look to see what’s different” between two states — for example, normal versus diseased, said Opiteck. “The spectra can enrich biomarker discovery. It can be combined with other strategies to provide biological insight,” he added.
MD Anderson’s Coombes said that looking at a panel of multiple biomarkers has more advantages than looking at single biomarkers. “One of the big pros is this: We’ve been doing single biomarkers for decades and there’s not a lot of them, so why not try something different?” he said.
However, Coombes cautioned that patterns are prone to false discoveries of significant profiles. “Validation is needed [to prove the significance of] patterns,” he said.
Leigh Anderson, the founder of the Plasma Proteome Institute, said that patterns are significant, even when proteins in the patterns have not been identified.
“This is reminiscent of the early days of 2D gels, when you couldn’t identify the spots, but you could see the pattern,” said Anderson.
Anderson said the reproducibility of patterns depends a lot of the manufacture of adsorptive surfaces. He agreed that before patterns can serve as significant biomarkers, they need to be validated.