The annual JavaOne developers’ conference may be an unlikely stomping ground for the head of informatics at a pharmaceutical company, but it’s where Bruce Ling gets some of his best ideas.
According to Ling, director of bioinformatics for South San Francisco-based Tularik, the life science informatics community has a lot to learn from mainstream software developers. “If you go to these enterprise solution conferences, they always use banks as examples because banking informatics is so rigorously controlled,” Ling said. “But to me, all these applications developed for life science research are no different than banking informatics — it’s just that we’re not as good as they are.”
Deciding that “research applications are just applications,” Ling set out to build a flexible, application-neutral framework to integrate Tularik’s bio-, chem-, and lead discovery informatics. Using standard enterprise architecture tools like UML (Unified Modeling Language), Rational Rose, and Together ControlCenter, Ling and his team built an enterprise discovery platform based on the traditional “model-view-controller” paradigm for designing interactive software.
The backbone for the platform is Sun’s Java 2 Enterprise Edition, which “gives you the capability to establish that kind of multiple-tier architecture,” Ling said. In addition to J2EE, Ling said that Tularik is also using Microsoft’s .NET technology for elements of the platform. Both J2EE and .NET allow developers to plug their applications into ready-made web services architectures to ensure interoperability, but there is a difference between the two: J2EE is language dependent but platform independent — developers must code in Java but their applications will run on Macs, PCs, Linux, or Unix systems; while .NET is language neutral but platform dependent — developers can code in C# or Java, but their applications will only run on Windows systems.
Because Tularik supports multiple platforms, J2EE made sense as the primary choice for the system, Ling said, except in those application areas where Windows is the primary operating system, such as structural rendering, where .NET is more effective. “I hope that the day comes soon when .NET runs on Linux,” Ling said, but until Bill Gates has a change of heart, Tularik runs two man application servers: one for J2EE and one for .NET.
The flexible architecture has proven to be a boon for Ling’s informatics team, which comprises only 10 of Tularik’s 400 employees, but supports the company’s bio-, chem-, lead discovery, and clinical informatics. “It’s a small group, but it’s very efficient,” he said. By following a “three zeroes” philosophy — “zero effort for development, zero effort for deployment, and zero effort for maintenance” — Ling said his informatics team has an edge over other pharmaceutical firms, where “the more they do, the more they create to maintain.”
The group relies on a mix of open source, commercial, and in-house-developed tools — a ratio that varies along the drug discovery pipeline. In bioinformatics, for example, Tularik relies primarily on publicly available and homegrown tools, Ling said, including an in-house microarray data management suite “that’s as good as any commercial software.” For cheminformatics, however, the company relies on commercial products from MDL, Accelrys, and Tripos.
In the area of lead discovery, Tularik recently turned to Panscopic, an enterprise software provider that is probably more familiar to attendees of JavaOne than of ISMB. Ling’s team added Panscopic’s J2EE-based software to retrieve and view data in its Oracle and XML databases, and for a range of visualization and data-mining tasks. For example, Ling said, Tularik was not happy with the visualization capabilities of its robotic readers for high-throughput screening, so it now transforms the reader data into XML and feeds it to Panscopic “so the scientists can monitor how the robot spits out the data at home.”
“It’s an enterprise application and we’re an enterprise,” said Ling of the decision to go with a general-purpose IT tool to handle Tularik’s research data. “We’re no different than people doing banking informatics…Everyone should develop software as robust as banking informatics. That’s where you want to put your money, right?”