NEW YORK (GenomeWeb) – With academic and industry developers of genomic-based tests for early cancer detection increasingly looking beyond circulating cell-free DNA, proteomic technologies could play a larger role in such efforts.
Prominent cfDNA researchers including Johns Hopkins' Bert Vogelstein have included protein level measurements as part of cancer detection assays they have published on recently. And last month, in perhaps the most notable instance to date of the cfDNA world building bridges to the proteomics space, cancer genomics firm Freenome announced a partnership with mass spec-based proteomics firm Biognosys to add protein data to Freenome's analyses as it works to develop its first commercially available screening test.
Freenome's interest in protein quantitation stems from its determination that identifying mutations in circulating tumor DNA will not by itself provide sufficient sensitivity for early cancer detection, said Imran Haque, the company's chief scientific officer.
"The problem is that in early stages of cancer, you get only a very, very small fraction of the circulating cell-free DNA coming from the tumor," Haque said, noting that this creates problems in terms of cost and the size of the blood draws required to detect cancer-linked mutations with high sensitivity.
"Even if you assume $1,000 genome sequencing costs, you still get numbers that are 10 to 100 times too high for [such assays] to really be viable commercially," he said.
Additionally, such assays would require impractically large blood draws due to the extremely low concentration of circulating tumor DNA present in early-stage patients, Haque said. "Unless you draw 150 mls of blood, there's a very strong chance that you will have literally zero mutated molecules in your sample."
Freenome presented these observations in a poster this year at the American Association for Cancer Research annual meeting. And, Haque noted, the company has for some time been skeptical that analysis of ctDNA alone would prove an effective route for early cancer detection.
"The approach that Freenome is taking is, broadly, to look at things [beyond ctDNA] like the role of the immune system and the rest of the body in giving you a signal about whether a tumor is present," he said.
The company is not alone in this approach. In fact, researchers and firms exploring genomic cancer early detection are largely in agreement that ctDNA-based testing will need to be bolstered by measurements of other analytes to provide sufficient performance. As a 2014 Science Translational Medicine study led by Johns Hopkins researchers demonstrated, ctDNA is not detectable in a substantial portion of both metastatic and localized tumors. And while liquid biopsy companies like Grail and Guardant Health have not necessarily highlighted these limitations to the extent Freenome has, both are looking to include analytes beyond ctDNA as part of their early detection assays.
With the recently announced Biognosys collaboration, however, Freenome making perhaps the most serious foray into proteomics of any genomic cancer detection outfit to date. The Vogelstein lab at Hopkins has included proteins as part of a ctDNA-based pancreatic cancer assay and in its multicancer detection CancerSEEK assay, but neither of those efforts involved a significant protein biomarker discovery component.
For instance, in their pancreatic cancer work, the researchers used four established cancer protein markers, including the longstanding pancreatic cancer markers CA19-9 and CEA. In the CancerSEEK assay, they cast a slightly wider net, using a literature search to identify 41 potential markers, which they then narrowed down to a final panel of eight via preliminary analyses of their expression in cancer patients compared to healthy controls.
The Freenome-Biognosys collaboration will use more extensive proteomic analyses to look for markers that could be useful in early cancer detection. Data-independent acquisition mass spectrometry is one of Biognosys' main areas of expertise and will be the primary technology the company uses in this work, said Biognosys CEO Oliver Rinner.
He said he believed DIA mass spec's high throughput and high reproducibility would be a good match for Freenome's AI-based approach to identifying biomarker patterns indicative of early-stage cancer.
"People have not been performing super large-scale, deep discovery proteomics experiments in the past because workflows were mainly tailored to going deep into a few samples," Rinner said. Recently developed DIA-MS workflows, however, allow researchers to reproducibly quantify thousands of proteins per samples using short liquid chromatography gradients, making proteomic experiments looking at thousands of samples more plausible.
"If you want to use AI approaches, you need lots of data," Rinner said. "And lots of data means lots of samples. So you need a method that provides depth and throughput at the same time. Because it only makes sense if you can go beyond thousands of samples."
Just as Freenome and others exploring genomic cancer detection have shifted their sights beyond ctDNA alone, proteomic cancer biomarker work has moved beyond a focus on markers secreted by tumor cells to looking for protein changes indicative of the patient response to cancer more broadly, Rinner noted.
"As history has shown, it's very hard to find a single, really good biomarker," he said. "And I think that is the interesting aspect of Freenome's work, that they are looking for signals beyond the circulating tumor [DNA], signals that are surrogate [tumor] markers. I think this is a very good fit for proteomics."
Haque said one reason Freenome believes proteins could be useful for early cancer detection is the fact that they are typically present at higher concentrations than ctDNA. Additionally, he said, while DNA usually enters the bloodstream only upon cell death, proteins are regularly secreted by cells in the course of their normal function.
"There are aspects of biology that you might be able to pick up on [by looking at] protein concentrations that you might not see through cell death processes alone," Haque said.
Of course, the potential advantages provided by relative abundance of circulating proteins compared to ctDNA are countered by the technical challenges of measuring these proteins. As is commonly noted in proteomic circles, while rare DNA species can be amplified for analysis using PCR, no comparable technique exists for proteins. This, along with the complexity and high dynamic range of samples like plasma, has limited researchers' ability to reproducibly quantify low-abundance molecules that might prove useful cancer biomarkers.
These issues are far from solved, but Haque suggested that technical improvements have made proteomic approaches more useful for the sort of cancer marker work Freenome is pursuing.
"I think [Biognosys] has interesting mass spectrometry technology that allows them to do a better job than some previous mass spec methods at quantifying a wide range of proteins and peptides," he said.
Rinner said that Biognosys has seen an uptick in interest in proteomics among genomic researchers and firms beyond Freenome, as well.
"We have several cases of customers who are interested in looking at the protein side after coming originally from the DNA side," he said. "I think people understand that the protein level is not the same as the DNA level, and so it makes sense to look at both."
Many genomics-focused researchers remain unaware of the advances proteomics has made, however, Rinner said.
"It's pretty clear to me that people either keep their eye on the DNA side or the protein side," he said. "They hardly ever talk to each other, and it's clear that many people [on the genomics side] have no idea what proteomics can do in the meantime. The perception is still that proteomics is at the stage we were at 10 years ago."
That perception is shifting somewhat, Rinner said.
"People have never doubted that proteins are important, but because [good protein data] was not available, it was not an option," he said. "Now people see more and more that it can be an option, and I certainly hope that this will have a kind of self-reinforcing effect where the more that this kind of data is used, the more people will become aware that you can connect to it to the genome or the transcriptome, as well."
AmirAli Talasaz, president and COO of liquid biopsy firm Guardant Health, said, however, that he believed proteomic techniques still needed further development. He added that while his firm agrees with the general consensus that ctDNA alone will not prove sufficient for early cancer detection, it has found that additional measurements of circulating nucleic acids significantly boost the performance of its assays.
"Just somatic alterations in circulating tumor DNA may not provide the clinical sensitivity that the market requires, but our belief is that if you combine that information with additional signals that are still nucleic acid-derived from liquid biopsies, you can have much higher sensitivity with very high specificity for the applications that we are talking about here in early cancer management and detection," he said.
Talasaz said that proteins "could be a huge potential opportunity," but that he believes proteomics remains hindered by a "lack of great analytical tools."
"One of the reasons people, including us, have made huge progress looking into different kinds of nucleic acid signatures is the fact that next-generation sequencing is a great analytical tool to really look at a lot of details of what's happening at the nucleic acid level," he said. "On the proteomics side… analytically, [researchers] are really limited. I believe that, if in the future, some analytical tools get developed for high-throughput proteomic analysis, there are a lot of opportunities that are going to be opened up. But currently we are tool-limited."