NEW YORK (GenomeWeb) – Despite high hopes and significant investment, plasma proteomics has thus far yielded little in the way of useful protein biomarkers. Only a handful of new protein markers and proteomic tests have made it to the clinic over the last decade, and even fewer have succeeded in achieving broad clinical adoption and commercial viability.
In a review published this month in Molecular Systems Biology, Max Planck Institute of Biochemistry researcher Matthias Mann and colleagues at Max Planck and the University Hospital at Munich's Ludwig Maximilian University suggested that technological limitations have been perhaps the most significant factor in hampering plasma protein biomarker efforts to date.
The researchers noted, however, that improvements in mass spec technology — especially in terms of throughput — are making new and potentially more promising approaches to protein biomarker discovery more feasible. In particular, they argued for what they termed a "rectangular" plasma protein biomarker development strategy, measuring on the order of thousands of proteins in large patient cohorts in both the discovery and validation phases of a biomarker project.
This contrasts with the "triangular" approach traditionally used in protein biomarker development, said Philipp Geyer, a graduate student in Mann's lab and first author on the paper. These experiments typically start off with a discovery phase in which researchers profile the proteomes of a relatively small cohort of cases and controls at great depth, measuring the expression levels of thousands of proteins.
Researchers then look for proteins with different levels of expression between cases and controls, and through either additional experiments or approaches like literature analyses, they arrive at a small set of proteins most likely to inform the clinical question they hope to address. They then typically move to a narrower analysis using either immunoassays or targeted mass spec to validate this subset of potential markers in larger numbers of patient samples — in the range of hundreds or thousands.
This approach, Geyer noted, has a number of potential pitfalls. Relying on a discovery process that measures thousands of analytes across a small number of samples creates the potential for overfitting and makes it difficult to distinguish between protein expression changes that truly reflect underlying biology versus those that are essentially due to chance.
Additionally, he said, such workflows require switching between several different technologies — moving, for instance, between a high-resolution mass spec instrument for discovery to a triple quadrupole for more targeted validation work to, possibly, an immunoassay format for additional validation and ultimate clinical implementation. Each type of assay comes with its own biases and limitations that must be accounted for, Geyer noted, adding that this can make it difficult for a single lab to successfully implement each step in such a process.
For instance, he said, while Mann's lab is a leader in shotgun proteomics, it has less experience in targeted protein assays.
"So, now we would have to learn another kind of proteomics, targeted proteomics," Geyer said. "And then the next step would be to establish an immunoassay, which is also different. How can I as a researcher who has never worked with antibodies now establish a really good immunoassay that has no bias and produces really reproducible data? That is really a problem."
Despite these issues, "triangular" biomarker development processes have become standard within plasma proteome work (and protein biomarker research in general) due to underlying technical limitations — foremost among them the difficulty of measuring large numbers of proteins in large sample cohorts. Traditionally, to achieve deep proteome coverage such experiments have required upfront depletion of high-abundance proteins, which is expensive and can bias measurements, or extensive fractionation, which significantly reduces assay throughput.
Geyer said, though, that improvements in mass spec technology and associated workflows have brought high-throughput, proteome-scale experiments almost within reach.
The goal, he said, would be to have discovery and validation cohorts in the range of hundreds to thousands of samples, and in all of them measuring un-depleted patient plasma at a depth of at least 1,500 to 2,000 proteins. This, Geyer said, would enable the aforementioned "rectangular" biomarker development process, in which data from multiple large, comprehensively profiled cohorts is used to identify potential markers.
Such a method would improve the statistical power of plasma biomarker discovery work, allowing for better identification of small but important changes, he and his co-authors wrote. The idea, they noted, is analogous to developments in genome-wide association studies, in which researchers "found that a joint analysis of as many samples as possible was superior to a sequential pipeline."
Mass spec technology is still not fully up to this task, Geyer said, but it is drawing closer. And, indeed, several recently launched large-scale projects point in this direction.
Last year, two sites, the University of Manchester's Stoller Biomarker Discovery Centre in the UK and the Australian Cancer Research Foundation International Centre for the Proteome of Cancer (ProCan) in Sydney, opened with the aim of implementing what mass spec vendor Sciex, which has relationships with both institutions, has termed "industrialized" proteomics.
ProCan plans to profile roughly 70,000 cancer tumor specimens over the next seven years, while the Stoller Center likewise aims to run thousands of patient samples. Both institutions are using Sciex instruments running the company's Swath data-independent acquisition mass spec method.
Another development that could push the field closer toward a "rectangular" model of biomarker discovery is the move of high-resolution instruments into the clinic. In a recent interview, Ravinder Singh, director of the Mayo Clinic Endrocrine Laboratory, noted that while his lab still relies mainly on triple quadrupoles, it has increased its use of high-resolution accurate mass instruments, which offer advantages in terms of accuracy and specificity.
Cost is currently a limiting factor, but Singh said that were money not an issue, he would prefer to do all his lab's clinical testing on HRAM instruments.
This is notable from the perspective of biomarker discovery in that HRAM instruments are able to collect discovery data on patient samples while simultaneously running a given clinical assay. Singh said that his lab gathers discovery data on a subset of patient samples and that while he is currently limited by data storage and analysis considerations, ideally he would like to run every sample "in a mode where I collect all the information on every sample."
Geyer said that this sort of approach is "exactly what [he and his co-authors] have in mind."
"If you have [HRAM] mass specs in the clinic, the mass spec has much more time than it needs [to measure the target analytes], so it can do thousands of [additional] measurements and you can collect much more information," he said.
"We always say to ourselves, imagine if you could profile [the plasma proteomes] of all the hospital patients in a city like Munich," he added. "You could get data on protein [levels] and combine that with information on their disease state or other clinical values in a database, and you could do so many things with that."