When medical personnel are combing through GBs of data to observe how medications affect DNA, they want something that’s simple, easy to use, and reliable. That’s what led Molecular Pathology Laboratory Network Inc. (MPL) in Maryville, TN, to develop and custom-build MPLNet, a portal through which physicians’ groups, hospitals, and other labs can access and exchange data and get medical test results. With MPLNet, clients and users will be able to track specimens and produce automated reports using a Web browser, once they’ve securely logged on from anywhere in the world.

Developing a need
MPL performs molecular-level testing for infectious diseases, cancers, and genetic abnormalities at the molecular level. Last year, the company processed more than 100,000 such analyses and this year promises even more, according to CIO Kenneth Billings. MPL needs to be able to handle resource-intensive DNA and RNA sequence analyses, as well as DNA micro arrays—data that first gets generated visually and is then read by a machine that translates it for diagnostic purposes. “We’ve got large data sets coming our way, and we needed a new system to handle them,” Billings said.

At present, most of MPL’s processes, functions, and tests don’t have huge volumes of system data. “We wanted a system that will handle gigabytes of data even if we don’t have that now,” Billings said. The human genome is an enormous 3 GBs of data, and if medical professionals start comparing 1 GB of sequence data against another 1 GB of sequence data, and images get added to the mix, the sequenced data grows exponentially. “We’re moving toward the threshold of terabytes of data and we will need to be able to handle that as it comes,” said Billings.

Product selection
After considering various options, including off-the-shelf solutions, for almost two years, Billings opted to mix and match individual products that he could tailor to MPL’s requirements. “We do testing that doesn’t lend itself to general laboratory reporting software packages,” Billings said. “A lot of our testing has images, graphs, and time-series data associated with it.” Billings also needed to accommodate multiple user platforms and operating systems, including Apple, Windows NT, and XP.

He decided to use Macromedia Inc.’s Flash product for his front end, because he thought it had richer visual impact for users accessing MPLNet. Billings has been working with the MX version of Flash and, despite some glitches early on, he’s pleased with the distinctive feel it gives MPLNet. “It allows us to do online testing and reporting in a very different way than Visual Basic LAS,” a more common, off-the-shelf product.

On the back end, he chose Caché 5 from InterSystems Corporation. Billings said Caché’s object-oriented database system is very fast and uses transaction bit map indexing. Caché 5 also includes full support of SOAP and WSDL, which enables applications to be built and Web services to be enabled very quickly, according to the vendor. Perhaps best of all is that Caché has extremely low development costs, Billings noted. “We’ve worked with them for a year testing it, and we’ll start paying when we have users—not developers—accessing the database. Right now, we’ve spent $150, and when we deploy we’ll pay $5,000 and it will go up from there. But as it goes up, it means we’ll have more clients, so it’s justifiable.”

MPL is also waiting for the soon-to-be-released Caché version for the Mac OS. “We use Mac Xserve for file serving needs and we’ve found it fantastically cheap with an unlimited license, so we’ll deploy on Xserve too. We think that will be extremely cost-effective, since we won’t have to pay the additional server fees,” he added.

Other sectors where this solution would work
The Flash-Caché hybrid isn’t just for healthcare and shouldn’t be dismissed as a technical curiosity for gene splicers, according to Carl Olofson, program director for information management and data integration software for International Data Corp., in Framingham, MA. “People who have been building applications with large data objects or with data-intensive processes have been pretty pleased doing this online with Caché,” he said. Olofson pointed to other sectors that could benefit, including scientific, technology, insurance, and law enforcement—anywhere that the level of interaction includes shoving around large files with historical data.

Regrets and relief
While MPL’s experience has been relatively painless, Billings said, if he had to do it over again, he would have hired a second software developer sooner. MPL’s growth in the last year has meant more pressure to get MPLNet up and running, he noted. Something he did do actually made a big difference during development and testing of MPLNet. “I’m glad we got a couple of people with different specializations—one’s good at back end with Caché and databases, and the other who’s good at front end, where you need creativity to make it interesting so people look at it and can navigate easily and quickly,” Billings explained. “It’s very hard to do both well.”