NEW YORK – A team led by researchers at the Max Planck Institute of Biochemistry and the University of Copenhagen has developed a set of proteomic classifiers for diagnosing alcohol-related liver disease (ALD).
Detailed in a paper published last week in Nature Medicine, the study used mass spectrometry-based tissue and plasma proteomics to identify plasma protein panels for detecting liver fibrosis, inflammation, and steatosis as well as predicting future liver-related events and all-cause mortality.
Notably, the mass spec-based panels performed as well or better than existing clinical tests for liver disease, said Matthias Mann, senior author on the paper as well as head of the department of proteomics and signal transduction at Max Planck and director of the proteomics program at the University of Copenhagen's Novo Nordisk Foundation Center for Protein Research.
This might seem like faint praise, but, as Mann observed, over its two-plus decades of existence, proteomics has struggled to live up to its potential as a tool for disease biomarker discovery and development.
"Proteomics people have been talking to themselves and saying how good [proteomics] is, but it has really never been shown that [proteomics-based tests] are actually as good or better than anything else that is out there classically," he said.
The field has faced a variety of challenges when it comes to diagnostics development. Traditionally, mass spec-based proteomics has had difficulty analyzing samples — particularly plasma — with both sufficient depth of coverage and throughput. Because of these limitations, biomarker discovery experiments have many times relied on relatively small cohorts and then shifted to larger cohorts for targeted validation of potential markers identified through the initial discovery work.
This approach has a number of potential pitfalls, though. For instance, 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.
Several years ago, Mann and his colleagues suggested in a review article in Molecular Systems Biology that the field should shift to what they termed a "rectangular" plasma protein biomarker development strategy, in which researchers do discovery work in hundreds to thousands of samples.
Recent developments in mass spec technology have made such an approach increasingly feasible. In the Nature Medicine study, the researchers used 21-minute liquid chromatography gradients and data-independent acquisition mass spec on a Thermo Fisher Scientific Orbitrap Exploris 480 instrument to characterize the plasma proteomes of 659 subjects, and 100-minute LC gradients to characterize the proteomes of liver biopsies from 79 subjects. They quantified a total of 5,515 proteins, 420 of those in both liver tissue and plasma. The total measurement time required was three weeks.
In the discovery set, the researchers looked at 459 subjects with ALD and 137 healthy controls. The validation set consisted of 63 subjects taken from an ALD screening study run by Odense University Hospital that looked at the effectiveness of transient elastography liver stiffness measurements for screening at-risk individuals and the general population for advanced liver fibrosis.
The researchers developed three different proteomic classifiers: a nine-protein panel for identifying significant fibrosis, a six-protein panel for identifying mild inflammatory activity, and a 12-protein panel for detecting liver steatosis. They compared the performance of these panels to 15 existing clinical tests, finding that the fibrosis and inflammatory panels produced better area under the receiver operating characteristic curve (ROC-AUC) values than any of the existing tests, while the steatosis test was outperformed by the controlled attenuation parameter (CAP) approach, which uses transient elastography (TE) to assess steatosis.
Depending on whether they were used to rule in or rule out a given condition, the proteomic panels outperformed some existing tests, while underperforming compared to others. For instance, the proteomic fibrosis test was better at ruling out significant fibrosis but worse at ruling in disease than TE or the enhanced liver fibrosis (ELF) blood test. Other existing tests were better than the proteomic model at ruling in fibrosis but worse at ruling it out. The same pattern held true for the inflammation panel.
The proteomic fibrosis test had the best performance of all tests for predicting liver-related events, with an AUC of .945 for predicting an event within three years and .933 within five years. It was also the top performer for predicting all-cause mortality, though its performance here was matched by several existing clinical measures.
Mann said that his group continues to work with its clinical collaborators at Odense University Hospital, noting that they were currently enrolling roughly 10,000 subjects for a liver disease study. He said enrollment would likely take until the end of the year and that his lab planned to profile the plasma proteomes of these individuals, with the goal of further validating and refining the liver disease signatures.
He suggested that a clinical version of the test could be useful for screening both high-risk populations and perhaps even the general population, noting that as much as 8 percent of the overall population is estimated to be at risk for liver disease.
While the performance of the ALD proteomics panels demonstrate that plasma proteomics can identify new markers that are comparable to or better than existing tests, Mann said further improvement in plasma proteomic workflows is needed.
He said that liver markers are something of "low-hanging fruit," given the organ's large size and its exposure to the circulatory system, both of which may make it easier to measure plasma proteins linked to liver disease than, for instance, plasma markers of early-stage cancer — an area where much protein biomarker discovery work has traditionally focused. He added that conditions like cardiovascular disease and metabolic diseases might also be more amenable to plasma protein biomarker discovery.
Ongoing advances in mass spec technology should also improve plasma biomarker discovery. Mann noted that in the time since the Nature Medicine work was done, the field has seen substantial advances. For instance, his lab has developed a number of workflows for Bruker's timsTOF line of mass spec instruments. At the American Society for Mass Spectrometry annual meeting in Minneapolis this week, Bruker introduced the latest version of the timsTOF, called timsTOF HT, which offers improved dynamic range compared to previous editions of the system.
The new instrument "should translate into better plasma coverage," Mann said. He added that he believed vendors including Thermo Fisher and Sciex have new instrumentation in the works that will also improve plasma proteome analyses.
Another company focused on boosting the depth and throughput of plasma proteome coverage is Seer, whose Proteograph system uses nanoparticle-based enrichment of proteins. The Redwood City, California-based company has signed commercial agreements with Bruker and Thermo Fisher to offer its system for sale with the timsTOF and Orbitrap Eclipse Tribrid and Exploris 480 instruments.
Mann said that the Novo Nordisk Foundation Center has a Proteograph platform that he and his colleagues are able to access. He said that the company's technology could be promising but that he was concerned about its per-sample costs, which he said could be substantial, particularly given the large numbers of samples that his team plans to run.
Near term, Mann said he hopes to achieve depth of coverage of around 1,000 proteins in plasma at a throughput of 100 or more samples per day.
"That's my internal goal," he said, and once it is achieved, "I think we would be in good shape."