Signals in the noise (Part 1 of 2)
Informatics challenges hinder progress in personalized medicine development
STORY PART 1 of 2
A lone technician sits before a computer monitor, watching avisual cacophony of images and data stream past her face, her retinas slowlybecoming inured to the assault on her senses—so much so, in fact, that it takesa moment for her to realize that the screen has suddenly stopped shifting and asingle form splays across the screen. Picking up the phone, she punches in anumber and waits for the other person to answer. They have the recommendationthey've been waiting for.
Jodie Foster in a scene from Carl Sagan's Contact? Possibly.
But it might also be any number of researchers andclinicians working in research hospitals and biopharmaceutical companies acrossthe country, as informatics advances push the capabilities of personalizedmedicine to the point where clinicians can get vital diagnostic and theranosticinformation in real time to assist them in making real-life decisions abouttheir patients, and effortlessly feed that information back to clinicalresearchers who are trying to develop the next generation of therapeutics.
As mentioned in the first half of our two-part series ontrends in personalized medicine ("A companion in your corner," ddn October 2012), regulators and pharmacompanies are increasingly pushing to develop diagnostic tests that designatewhat patients are best suited to receive what drugs, whether due to improvedefficacy or safety. The canonical example is, of course, the HER2 biomarker andits use in determining whether a breast cancer patient should receive a druglike Herceptin that specifically targets Her2-positive tumors.
Biomarker signatures that only involve one or two genes,however, really only give us so much of an edge. If that biomarker is found in40 percent of patients tested, that is still a lot of patients, and this merelytells you it has a better chance of working in Jessica, who also carries thatbiomarker.
Without minimizing the work that went into identifying andvalidating those single biomarkers, these tests are only really part of theequation—the low-hanging fruit, if you will. It's the equivalent offingerprinting a suspect on the basis of one line of a whorl in one quadrant ofthe thumb. Fingerprints are much more complex, which is why they can be used toidentify individuals.
"A single biomarker will only give you so much insight, andas we progress to more complex diseases, combination signatures becomeincreasingly important," echoes Ger Brophy, general manager of new productdevelopment at GE Healthcare in Fairfield, Conn.
As an example, at the Markers in Cancer Meeting in October,Roland Kappler and colleagues from the Dr. Von Hauner Children's Hospital inMunich described their efforts to identify more complex genetic signatures asprognostic markers in hepatoblastoma in children. Examining the methylationpatterns of the RASSF1 gene and correlating those patterns with clinical data,the researchers were able to identify a distinct epigenetic pattern thattightly linked with the likelihood of metastasis and overall survival rates inpatients.
But genomics is only one part of the biomarker equation,albeit possibly the most linear when it comes to linking biomarker withoutcome. As we begin to expand our appreciation of an individual's biomedicalfingerprint, it will likely be necessary to expand into other methodologies.This move has started on some fronts.
In April, Research Triangle Park, N.C.-based Metabolonannounced a research initiative with Osaka, Japan's Takeda Pharmaceutical toidentify novel therapeutics and biomarkers using its metabolomics expertise.And in August, the company announced its completion of the acquisition oflipid-metabolism specialist Lipomics Technologies, further expanding itstechnology base.
"We are the leader in the commercialization of metabolomics,with a profitable commercial life-sciences service business, and have launchedthe world's first metabolomics-based diagnostic test for type 2 diabetes riskbased on biomarkers that measure insulin resistance," said Metabolon CEO Dr.John Ryals in announcing the acquisition. "We expect that in 2013, we will bemarketing additional diagnostic products aimed at diseases related to obesityand cancer, and are committed to maintaining our position as the world's leaderin metabolomics."
Companies like Ezose Sciences, based in Pine Brook, N.J.,meanwhile, are attempting to leverage expertise in glycomics and the company'sGlycanMap platform as the basis of a diagnostics pipeline. In April, thecompany announced a collaboration with Hirosaki University to identifypotential glycomic biomarkers to predict and monitor prostate and otherurological cancers.
And even cellomics has started to make inroads into thebiomarker world. Also presenting at the Markers in Cancer Meeting, researchersfrom Durham, N.C.-based Argos Therapeutics discussed efforts to identifycell-surface signatures for patient response to its immunotherapy candidateAGS-003 in patients with metastatic renal cell carcinoma.
Argos researchers applied informatics methodology morecommonly used in microarray analysis to flow cytometry of cytotoxic T cells andidentified combinatorial expression patterns of cell surface markers thatcorrelated with superior outcomes in patients receiving AGS-003 and sunitinib.By parsing the data even further, they were able to identify other biomarkersignatures, which they described as markers of immune function, or MIFs, andtightly correlated with progression-free survival and overall survival outcomesin the study patients.
Driven by data
The inherent complexity of these signatures, however,increases the need for informatics solutions to help distinguish the signalfrom the noise in identifying true signatures. Unlike the HER2 tests, we are nolonger looking at a simple binary yes-no answer.
Researchers aren't just limiting themselves to 'omics data,either.
"While the in-vitrosignature—e.g., genetic biomarkers—isimportant, the in-vivo stage—e.g., imaging—is also very important togetting a complete answer," says Brophy. "The question then becomes how do youcombine data from a variety of sources such as electronic medical records(EMRs), pathology, radiology, etc."
For GE Healthcare—which has extensive expertise in medicalimaging methodologies such as PET, MRI and CT scans—part of the answer came inmoving that expertise into the world of pathology.
"The digitization of radiology revolutionized that field,and we feel the time is now right to digitize pathology and take it from asubjective art to an objective science," Brophy adds. "This would give pharma aquantitative tool to monitor the performance of a drug on pathology slides,rather than rely solely on the qualitative analysis of a pathologist."
To that end, GE Healthcare recently launched Omnyx, a jointventure with the University of Pittsburgh Medical Center (UPMC). According toBrophy, UPMC brings an extensive collection of annotated tissue slides andoffers Omnyx access to a constant influx of thousands of patients each year.The focus is on cancer diagnostics and developing systems where clinicians andpathologists can store, retrieve, annotate and share data easily and quicklywith colleagues.
"While the scanner is the most obvious component, thesoftware interface and back end is key to the system," he adds. "And of course,the challenge is in not just developing a tool that reflects what thepathologist is doing now, but also how they will be working in the future."
"Clinical diagnostics are not new—people have beenperforming urine tests or blood work for years," adds Trish Meek, director oflife-sciences product strategy at Waltham, Mass.-based Thermo Fisher ScientificInc. "The challenges of molecular diagnostics, however, are quite unique asresearchers try to perform standard assays, while always incorporating newtechnologies and techniques into their practice."
As its clients' needs evolved, Thermo Fisher Scientificextended its knowledge with lab information systems (LIS) and lab informationmanagement systems (LIMS) into helping pharma customers with biomarkeridentification and validation—and ultimately into research hospitals.
"We work with the customer to take them beyond thetraditional LIS, so our job is still about operations on one side, but it isalso about making sure the information is clear and actionable on the other,"says Meek. "That's why we developed the web-based interface for the end user."
As biomarker signatures become more complex, pharmacompanies are likely to rely more heavily on their diagnostic partners and canexpect a significant shift in the clinical trials landscape.
"Biomarker validation will become an increasingly importantpart of the clinical trial process," Brophy offers. "As pharma companiesapproach the Phase IIb trial of their new drug without a marker, they may beginto get worried that they don't have a companion diagnostic."
GE Healthcare's 2010 acquisition of molecular diagnosticsspecialist Clarient, which is based in Aliso Viejo, Calif., was a step towardaddressing this need.
"GE and Clarient are in a number of discussions with pharmacompanies as to how we can help, leveraging our CLIA-certified labs to identify,develop and validate these diagnostic resources," adds Brophy.