Skip to main content
Premium Trial:

Request an Annual Quote

Lund Study Highlights Difficulties Facing Alzheimer's Plasma Protein Biomarker Research


By Adam Bonislawski

Scientists from Lund University have been unable to validate a plasma protein biomarker panel that was previously identified as having links to Alzheimer's and was explored as a potential commercial diagnostic for the disease.

The research, published last week in PLoS One, highlights the continuing challenges facing plasma protein biomarker discovery for Alzheimer's disease – specifically, the difficulties of doing biomarker discovery in limited sample sets and the risks in investigating markers not obviously linked to the biology of the disease in question, Oskar Hansson, a Lund University researcher and co-author on the paper, told ProteoMonitor.

"When you study biomarkers that are not obviously involved in the pathogenesis of the disease you are studying – as in this example, where none of these [proteins] have really been shown to be involved in what is happening in the brain – then you need to be cautious," he said.

In the study, the researchers evaluated a panel of 18 plasma proteins previously identified as Alzheimer's biomarkers in a 2007 Nature Medicine study. The Lund team was unable to reproduce the original study's results and found that, according to its new analysis, none of the 18 proteins could identify Alzheimer's patients with high diagnostic precision.

While the original study claimed the 18 markers could classify Alzheimer's patients with accuracy of close to 90 percent and identify patients that would progress from mild cognitive impairment to Alzheimer's, the PLoS One study found the markers identified Alzheimer's patients with an area under the curve of only 63 percent.

Stanford University researcher Tony Wyss-Coray, one of the authors of the original Nature Medicine paper, agreed with Hansson, telling ProteoMonitor that "measuring plasma proteins is a huge challenge, and I think most people who have experience in the field recognize that."

Calling the Lund effort "a very nice study" with "nothing to criticize," he said that, looking back on the 2007 work, he and his collaborators had been "probably a bit naïve and too enthusiastic" about the markers they identified.

"I think there are differences [in the plasma proteome of Alzheimer's patients], and actually we continue to see differences, but I think there are a lot of issues," he noted.

Wyss-Coray said that his lab has since moved away from blind biomarker discovery work and is focusing on identifying signaling pathways linked to Alzheimer's with the aim of better understanding mechanisms of the disease.

"I don't think you have to throw out proteomics discovery entirely," he said, "but you have to do the biology afterwards."

Wyss-Coray compared the current state of plasma protein biomarker discovery in Alzheimer's to the early days of genomics when researchers were putting out a steady stream of studies trumpeting discoveries of potentially important SNPs.

"You basically saw one paper after another finding a SNP linked to a disease, and most of them weren't reproduced in independent datasets," he said. Small sample size, he said, was a fundamental problem with these studies and, likewise, is a key issue with proteomics-based discovery studies today.

"Most studies deal with sample sizes that are just not realistic to produce reproducible results," he said, noting that the initial study looked at roughly 100 patients and controls. The PLoS One study, meanwhile, examined 174 controls, 142 Alzheimer's patients, and 88 patients with other dementias.

Wyss-Coray said he thought that for undirected plasma biomarker discovery to be effective, researchers would need to look not at hundreds of samples but tens of thousands.

"Just measuring proteins in cases and controls and then thinking you have a biomarker for the disease isn't going to work unless you have the luxury of measuring 10,000 samples," he said.

Based on the results of the Nature Medicine paper, Wyss-Coray founded the biotech firm Satoris, which aimed to commercialize the markers. The company was purchased by Rules-Based Medicine – now Myriad RBM – in May of last year (PM 5/17/2011). The proteins have made it into research panels now owned by Myriad RBM, but not into an FDA-approved diagnostic, and, Wyss-Coray said, he believes work on the markers is "on the backburner."

Alzheimer's has been a prime area of focus for protein biomarker research. Last June, a report commissioned by proteomics firm Proteome Sciences predicted that protein biomarkers for Alzheimer's disease will represent a cumulative $9 billion market over the next ten years (PM 6/3/2011). In addition to their promise for diagnosis and predicting progression of the disease, early detection and progression biomarkers are also expected to play an increasing role in pharmaceutical research into drugs for the disease, helping with selection of patients for clinical trials and monitoring response to therapy.

While the bulk of Alzheimer's biomarker work has looked for markers in cerebrospinal fluid, plasma markers are desirable because they could be tested for without having to perform a lumbar puncture – making them more useful for purposes like regular monitoring.

Holly Soares, director of clinical neuroscience biomarkers at Bristol-Myers Squibb, agreed with Wyss-Coray that much larger sample sizes are needed if research into plasma protein markers for Alzheimer's is to be effective. She was more optimistic, however, about the possibility of putting such studies together. Soares, who is chair of the Alzheimer's Disease Neuroimaging Initiative Biomarkers Consortium project team, said that it has recently become feasible to put together a biomarker discovery study on the scale of 10,000 samples.

"I think many of the large Alzheimer's research centers routinely collect plasma, so I think it's very reasonable to expect that we'll be able to combine some of those cohorts to address these questions in a meaningful way," she told ProteoMonitor, suggesting that such an effort could emerge at this summer's Alzheimer's Association International Conference.

"There we'll be talking about all the reports on blood-based biomarkers and many of the major players with large cohorts are going to be there, and that might be the place where we think about combining our datasets."

Soares noted that, despite discouraging results like those presented by the Lund group, researchers have seen "some promising reproducibility [in Alzheimer's plasma markers] across some large cohorts."

ADNI, the Australian Imaging, Biomarker & Lifestyle Flagship Study of Ageing, and the Texas Alzheimer's Research Consortium have each recently screened samples against discovery panels from Myriad-RBM, she said. "So altogether, that's about 2,500 samples that have been looked at, and there is some reproducibility."

"We are seeing pretty consistent changes across very large cohorts, and there does seem to be some association with the genotype, which is promising," Soares said. "So I think that we might finally be seeing something that could be usable, at least as a screening tool to get us to the next step."

She said ADNI's plasma biomarker work is currently in an "intermediate stage" where the research platforms being used still "have some batch-to-batch variability."

"What we would like to do is have a bit more of a robust assay that didn't have this variability and then do some more testing of the samples to make sure what we're finding … can be used," she noted.

Soares also said the group might look to add to ADNI's plasma biomarker data, building on its earlier work using RBM's panel to measure 189 proteins in subject blood samples taken at baseline and one year later (PM 12/10/2010).

She noted that there is currently "a proposal on the table" to combine some of the more promising markers from this work into a smaller panel, reanalyze them, and, "if the data looks good," go back to ADNI for more funding to continue monitoring these subjects.

Have topics you'd like to see covered in ProteoMonitor? Contact the editor at abonislawski [at] genomeweb [.] com.