A recent addition to New Haven’s biotech cluster, TurboGenomics is making friends quickly. Genaissance Pharmaceuticals, also based in New Haven, is the company’s first customer, licensing the startup’s TurboBlast software, a high-performance version of Blast.
The newly launched software is the first offering in a full bioinformatics platform that TurboGenomics plans to release by the middle of 2001.
The 10-employee company is founded on a proven technology base, according to CEO Greg Gardiner, a former Pfizer executive and recently retired head of Yale University’s office of cooperative research. Both TurboBlast and the upcoming platform — tagged with the working name of TurboBench for now — are variants of parallel and distributed computing technologies licensed from TurboGenomics’ parent company, Scientific Computing Associates.
Gardiner told BioInform that this is the fourth iteration of this distributed computing technology that has already been successfully applied to financial analysis, oil exploration, and ship design. “This product presents no technical risk,” Gardiner said.
In addition to the licensing agreement with Genaissance, TurboGenomics has five or six other ongoing projects, including a custom deployment of the platform for another New Haven neighbor, Cellular Genomics.
TurboGenomics is also ready to close a $5 million round of financing from Trautman Wasserman & Co. Greg Trautman, president of the investment banking firm, commented, “We’re impressed with TurboGenomics’ platform.” Trautman sees the company’s technology as a key component in the growing bioinformatics sector, “which should see substantial growth within five to 10 years.”
Gardiner expects the company to go public within two to three years.
TurboBlast and TurboBench stem from Scientific Computing’s Paradise technology, which can be used to create high-availability ensemble applications from otherwise independent parallel, distributed, and sequential programs running within a network. Specifically, TurboBlast maximizes the performance of any version of Blast by dividing the task among a group of networked computers, getting an additional boost from idle cycles on desktop PCs in the network. The proposed TurboBench platform integrates a user-defined range of applications, databases, or search engines into a seamless computing infrastructure that can be customized through a user interface.
The execution engine at the heart of the TurboBench platform —labeled with the working name of TurboHub — executes a given specification using a variety of networked machines. “It does it not by pushing the work out,” said TurboGenomics founder Andrew Sherman, “but rather by simply posting information about what work needs to be done and letting the machines grab work from the remote locations.” The TurboHub runs on Windows, Unix, or Linux, but the applications run wherever they are located.
The TurboBench platform also draws from technology developed by sister company Metaserver, another Scientific Computer Associates spin-off, to effectively link the user interface and the chosen suite of applications to the TurboHub and to each other.
Originally developed to integrate e-commerce applications, the Metaserver technology supplies “wrappers” or “conduits” for each application so that they can communicate with each other through the TurboHub. Sherman described the conduits as jigsaw puzzle pieces that enable the individual components to plug into the execution engine.
It’s possible to write a conduit for any application that may become part of the system. TurboGenomics intends to write some generic conduits — a conduit that will use SQL to talk to any relational database, for example — but will also publish the interface and specification information that is required for anyone to write a conduit for any application.
The company plans to post this information on its website and through other means, possibly as early as the first quarter of 2001.
Conduits may also be written for the front end of the system, so a user may opt to use a favorite desktop interface rather than the visual programming environment that TurboGenomics is developing for the system.
Sherman cited an example of a biologist taking accession numbers from an Excel spreadsheet, plugging them into an Internet-based Blast routine, and then running the results through a Corba-based analysis program. “In the classic view of bioinformatics, the biologist would cut and paste the results of one screen into another,” Sherman said, “which really wastes a lot of time.”
Using TurboBench, the biologist could build a custom processing pipeline for the desired analysis steps through the user interface. The user interface would communicate via a conduit with the TurboHub execution engine, which would then find the appropriate machine to run each task.
As each application is chosen and run, it could communicate with the other components through the conduits plugged into the TurboHub, allowing Common Object Model-based applications to talk to Java or Corba applications. The TurboHub can execute on a network, over the Internet, or over an organization’s intranet using standard networking communication technology.
The conduits are written in “a little bit of Java code,” Sherman said. “It’s not a tremendously complicated process, but you have to know something about programming and you have to know something about Java and that is a skill set that some biologists have but many don’t.”
Sherman stressed that the technology is focused on application integration rather than data integration, which he acknowledged as a key issue in the bioinformatics field. TurboGenomics is actively partnering with people to address the data integration problem, Sherman said.
The entire TurboBench platform will comprise the TurboHub execution engine, along with the necessary conduits for a given number of front- and back-end components. Some will come with the system and others will be options. In addition, any user will be able to write conduits to plug in particular applications.
TurboBlast is a standalone system, but can also plug into the platform through a conduit. The company is applying the TurboBlast technology to other applications, and is currently working on a “Turbo-charged” version of a Smith Waterman sequence analysis that is still in early-stage internal testing.
The company believes its technology is well-suited for partnerships. Even seeming competitors, such as InforMax or Lion, which have enterprise-wide systems, are potential partners, according to Sherman, since they offer products that could be implemented in the TurboBench platform.
TurboGenomics expects to have several TurboBench prototypes delivered by the first quarter of 2001. These will feature the TurboHub environment and a limited set of conduits.
“The heart of our market is the biotech companies,” said Sherman, pointing out that the technology addresses the underlying IT issues that these companies face, allowing them to hire biologists rather than computer scientists.
“The Turbo platform will run from day one on an annual license basis that will cost less than a computer expert,” said Gardiner. “It lets biologists do bioinformatics their way,” added Sherman.