Joining the rapidly growing ranks of those seeking to develop a protein pattern-based cancer diagnostic, the Wistar Institute of Philadelphia and local bioinformatics start-up Cira Discovery Sciences announced this week that they will collaborate to look in serum samples for patterns indicative of several types of cancer.
David Speicher, who will lead the proteomics data collection and analysis effort, said his lab could possibly produce its first diagnostic test — which will be an immunoassay — within two-and-a-half years, but that four or five years was a more realistic time frame. Speicher said that his lab is currently developing the necessary technology and collaborations for mass spec-based discovery, but that he hopes “to have a reasonably reliable system set up within four to six months, that we will then be able to start applying seriously to samples.” Speicher is a professor and the systems biology division chair at Wistar (see PM 12-12-03).
Speicher is not looking at this point to join the race to produce an ovarian cancer test (see PM 2-20-04), although he said that ovarian was one of the cancers “on the top of our list” for biomarker study, along with breast, colon, lung, skin, and pancreatic cancer. He said that he has already set up collaborations with local clinicians in Philadelphia to study breast cancer, lung cancer, and melanoma, and so will likely start with those. He is still seeking collaborators for the other cancers of interest. “If you have a good technology that’s set up and working, one could apply it to all of these diseases,” Speicher said. He added that while current collaborations have been local, he anticipated that casting a wider net would be necessary later because “in some cases perhaps we may not be able to get as many samples as we’d like.”
Speicher said his group’s approach, which he described as “SELDI-like,” would emphasize the advantages of SELDI — such as its ability to be very high-throughput — while “work[ing] around the disadvantages,” such as its low resolution, low mass accuracy, and difficulty in identifying peaks, he said. “One big problem with SELDI is … when you see a feature in a pattern, to track down and verify what that pattern is is very difficult if not impossible,” since the starting amount is small, and scaled up models do not always track the same component, he said. Instead of using SELDI, Speicher is looking into “using similar types of general chemistries that SELDI provides on the substrate, but using those on conventional beads or modified membrane formats,” he said. He will then produce patterns of peaks with an undecided mass spec platform; select significant patterns using Cira’s algorithm; identify the proteins of interest; and then use the identified proteins to construct an immunoassay test that can be used as a diagnostic. Speicher expects this last part to be the main bottleneck in the process, depending on how many significant markers make up the patterns he finds. “Certainly if we come up with a handful, that’s not going to be too hard to produce immunological reagents, but if we discover 40 or 50 we want to analyze, that’s going to be much more problematic,” he said.
Cira has developed pattern recognition software along the lines of the type Correlogic used in its collaboration with the NCI-FDA (see story, p. 1) but with a different data mining approach. Joining Cira made sense to Speicher because he felt patterns are the way to go.
Wade Rogers, president and CEO of Cira, said that his pattern recognition algorithm was more “exhaustive and complete” than other types of algorithms — such as the one that Correlogic uses — that depend on a priori assumptions to reduce the complexity of the search. Cira reduces the complexity only after the exhaustive search is complete, by creating a “polynomial solution” that can provide complete coverage of all possible patterns “in a reasonable amount of computing time,” Rogers told ProteoMonitor’s sister publication, BioInform.
Rather than competing with other protein pattern discovery efforts, Speicher said he hoped to simply add to the information available. “Since it’s not clear what technologies are the best, it’s a good idea to take multiple approaches,” he said. Speicher did note, however, that he was looking to address certain criticisms that have come up in regards to Petricoin and Liotta’s original ovarian cancer studies — in particular, the issue of artifacts that may result from sample handling, and the possibility of proteolysis occurring when samples sit at room temperature. “We’ve seen, in some very preliminary analyses on acute lung injury samples, substantial differences in the major protein patterns, some of which are probably due to proteolysis,” he said.