This article has been updated from a previous version, which incorrectly identified the manufacturer of Geneva Bioinformatics' Phenyx software.
Proteome Software, a Portland, Ore.-based bioinformatics startup, is hoping potential customers share its belief that you can't have too much of a good thing when it comes to proteomics search engines.
The firm's first product, Scaffold, compares results from multiple mass spec search tools to calculate a probability score that gives researchers more confidence in their protein identifications. First, Scaffold takes as input data from Matrix Science's Mascot software and Thermo's Sequest software, and compares them to each other or to X! Tandem, an open source protein search algorithm that comes as part of the package.
In addition, Proteome Software has implemented its own version of the Institute for Systems Biology's Peptide Prophet and Protein Prophet statistical algorithms within Scaffold to assign probabilities to the results from each search engine, as well as the combined results, to provide an overall probability score for each of the proteins identified.
"The biggest thing that you're going to run into with these types of searches is a lack of confidence in the actual identification," said Mark Pitman, sales and marketing director at Proteome Software. "There are a lot of false positives generated by any one of them alone. So the biggest hurdle that you're going to face is trying to say, "Okay, I have these proteins listed from Sequest, but how many of them are really there?"
"The biggest thing that you're going to run into with these types of searches is a lack of confidence in the actual identification,"
As proof of the uncertainty of using any one method alone, Proteome Software used its approach to compare a control sample of 10 known proteins using Sequest, Mascot, and X! Tandem. Only 34 percent of the total peptides were identified by all three programs, a finding that was "actually a lot less than we had assumed," Pitman said. In addition, 9 percent of the peptides were identified only by Sequest, 7 percent only by Mascot, and 19 percent only by X! Tandem.
With Scaffold, Pitman said, "You can actually get a really good idea of whether a protein is there. If it's highly scored by three different algorithms, you can be fairly confident that that protein was actually in that mix." On the other hand, he said, "If any one of them identified it just by itself, you might want to dig a little bit deeper, take a look at the fragmentation of that spectra, and say, 'Okay, does that make sense from a chemistry basis that that would actually look like that if it split up inside of a mass spectrometer?'"
Scaffold also provides data-visualization and data-management capabilities, Pitman said, allowing researchers to compare proteins from several samples and store data from multiple experiments in a single file format.
Pitman stressed that Proteome Software considers Scaffold to be an add-on to current protein search engines like Sequest and Mascot, and not a competing product. "One of the big things that we wanted to do at the outset was not be a competitor to them," he said. "They already have their field, and almost everybody is going to have one or the other of those in their lab already."
Pitman said that the company is also working on adding support for other packages, such as Agilent's Spectrum Mill and ProteinLynx from Waters, to Scaffold.
The closest competition for the company may come from Bruker Daltonics' ProteinScape, a package it co-developed with proteomics firm Protagen that can integrate data from Mascot, Geneva Bioinformatics' Phenyx, and Genomic Solutions' ProFound. Pitman acknowledged that ProteinScape is "similar to what we're offering," but added, "They don't apply the same type of probabilities that we do to their search algorithm."
Proteome Software was founded last August by Mark Turner, Ashley McCormack, and Brian Searle — all researchers at the Clinical Genomics and Proteomics Program at Oregon Health & Science University. Pitman and another founder, James Brundage, joined the firm shortly afterwards.
Scaffold debuted at the American Society for Mass Spectrometry conference in San Antonio earlier this month, and Pitman said the feedback from attendees was very positive. "We talked to a lot of the target market we were going after," he said. In addition, "We were able to make some tentative alliances with some of the machine vendors. They wanted to take our software and promote it to their clients as another tool."
Pitman did not disclose which instrumentation vendors the company was in discussions with.
Proteome Software has seven beta-test sites for Scaffold so far — a mix of pharma and academic institutes, Pitman said — but no paying customers yet.
Scaffold is priced at $5,000 per seat for an academic license, and $6,000 per seat for a commercial license, with discounts available after five seats. Pitman said that the company is also considering site licensing and other models that might appeal to the core labs that it has identified as its most likely short-term customer base.
"One of the questions we would like to answer for ourselves is whether a biologist who is interested in proteomics but not necessarily a bioinformaticist themself, whether or not this would make sense to them, but we just don't have a data point on that," he said.
— Bernadette Toner ([email protected])