SAN DIEGO — “This is the counterrevolution against systems biology, and it’s thin slicing,” Francis Barany, keynote speaker at the Feb. 17 Clinical Genomics meeting, told Pharmacogenomics Reporter last week.
By that comment, Barany, a professor of Microbiology and Immunology at Cornell University’s Weill College of Medicine, summarized what he sees as a step critical to getting molecular diagnostics for tumors into the clinic and the minds of doctors. The term “thin slicing,” popularized by author Malcolm Gladwell, and which Gladwell credits to the field of psychology, refers to the ability of humans to grasp their environment using a tiny amount of information. In the context of Barany’s comments, it refers to the paring down of microarray expression data to its barest essentials. Just as importantly, physicians need a reliable and cheap mutation-detection platform, Barany said.
A key part of the argument is that doctors really have little use for the deluge of information that arrays provide. That extra data can even work to prevent molecular diagnostics from entering the clinic, said Barany. “People complain that physicians are too conservative. It’s not true. They won’t allow new information into the clinic that doesn’t help the patient — you can’t treat a 30,000-gene expression profile,” he said. The cost of microarrays is another major barrier, he added.
To be sure, Barany has an interest in promoting a simplified approach. He is the director of an international colon cancer research consortium that has been developing mathematical approaches to separate useful expression data from the chaff. A novel universal DNA array he and collaborators developed — featuring PCR-ligase detection reaction and endonuclease V-ligase mutation scanning — should “allow clinicians to use simpler and more cost-effective techniques … to verify the clinical significance of various markers,” he said.
In tumor research Barany presented at the conference, his group used expression arrays to identify a small subset of important transcripts using a mathematical model. With the PCR-LDR/endo V-ligase mutation scanning method to find mutations in these genes, coupled with methylation detection, the group boasts 99 percent sensitivity for mutation detection, compared to 62 percent for automated sequencing and 92 percent for sequencing using manual reading, he said.
Ultimately, the plan is to develop sensitive, point-of-care devices for physicians to identify tumor-related mutations on site, said Barany. Applied Biosystems, which has an exclusive license for PCR-LDR, would have to decide to commercialize such a product, he said. Before that can happen, clinical diagnostic laboratories would need to adopt the method as a homebrew, he added.
“The advantage of the universal array coupled with LDR is flexibility,” said Barany in an e-mail to Pharmacogenomics Reporter. “By carefully designing oligomers and reaction conditions, the reusable arrays can be ‘programmed’ to identify different mismatches in different runs.”
The Adoption Question
But are major labs interested in expression microarrays? Quest Diagnostics, for one, has yet to adopt them. After a conversation with Barany about his system, Monsoor Mohammad, the director of microarray technology and proteomics at Quest, said his company is examining “all these claims” to figure out which microarray platform has “the kind of robustness to get into the clinic.” The qualities most important to the transition of microarrays into the clinic are robustness, clinical relevance, time efficiency, and cost, Mohammad said.
Clinical laboratories have a strong interest in sticking with what works. “They love PCR-based tests, because they’re very familiar with them, there are already CPT codes for getting reimbursed with them, and all that has yet to be worked out on the microarrays,” said Bob Schuerer, chief operating officer of Arcturus Biosciences. In microarray tests, “there is not much of a menu yet — there are a lot of PCR-based tests out there now. No one has really come up with the large slots of genes for diagnostic purposes, other than the one that we’ve come up with,” he said.
That test is Arcturus’ cancer-of-unknown-primary — or CUP — microarray profile, which the company recently [see BioArray News, 2/9/05] licensed to USLabs and Netherlands-based Agendia Biosciences, and “requires the use of many thousands of genes,” Schuerer said. The CUP test is used to classify tumors of a type that occur in patients about 100,000 times a year in the United States, he said.
US Labs offers the test, which it calls TUO — or tumor of unknown origin — for $2,000, and for research purposes only, according to the company’s website.
More Thin Slicing
But Arcturus still sees PCR as a reliable way to get platforms into clinical labs, said Schuerer.
Indeed, Mark Erlander, Arcturus’ chief scientific officer, presented research at the Clinical Genomics conference that Barany cited as an excellent example of thin slicing in action. Starting with discovery experiments using large-scale expression microarrays, Arcturus eventually identified two genes, HOX 13 and IL 17, correlated with the risk of breast cancer recurrence following tamoxifen treatment, Erlander said during his presentation. The ratio of HOX 13 to IL 17 can predict up to a seven-fold risk of recurrence, down to a 97.5 percent chance of no recurrence, Erlander said.
The company has licensed the PCR signatures out to Agendia and US Labs as the MammaPrint diagnostic, said Shuerer. “We are in active discussions with other major labs,” who have shown “very high” interest, he added.
But then, perhaps a slow shift from giant expression microarrays to smaller diagnostics is just the natural development of the field after all. “Systems biology and thin slicing move in concert,” said Dan Clutter, vice president of sales at Nimblegen Systems. “You do want to have the context of the cell, for the systems biology approach, but you also have to deal with smaller pieces of data, because you're swamped with data.”
While microarrays are cumbersome, they may find their place in the larger scheme of things, said Clutter. “Doctors basically used to give you something to see if it works, then give you something else — to some extent that still works today,” he said. “But we're hoping to get to a point where we can literally scan your genome and find out which medicines will work for you and which won't.”