LowBy M.L. Baker | Posted 09-08-2005
Drug Scientists Slow to Adopt Open Source
A surge in open-source technology offers many businesses a potent combination of power and low cost, but the pharmaceutical industry remains well behind the curve.
That was the message at a session on data visualization and integration at a semiannual meeting of the nation's chemists in Washington last week.
No one thinks open source can fill everyone's software needs, but a "forceful" open-source community could create modules that bridge gaps in existing software designed for the pharmaceutical industry and help it become more efficient, said presenters. Yet so far, few are stepping up to the plate.
Interest in visualization software has surged in the pharmaceutical industry in the past couple years, but interest in open source has not, said Carol Rozwell, vice president of life sciences at consultancy Gartner Inc., who did not attend the session. However, she said, interest might pick up as the pharmaceutical industry sees IT more firmly integrated into drug discovery and development.
In a recent survey, Rozwell's group looked at all sorts of software used by biopharmaceutical companies to find out if it was developed in-house or purchased from a software vendor.
Overall, there was an increasing tendency to use off-the-shelf products. "The one holdout area," she said, "was discovery. Many clients express concern that the commercial products don't quite fit the bill, hence the tendency to believe that the only option is to develop their own software."
Software can come from pharmaceutical companies' internal efforts, commercial software vendors, academics and open-source communities.
Each has its place, and its pitfalls, said Warren Delano, head of Delano Scientific, who presented his open-source product PyMOL at the semiannual meeting of the American Chemical Society.
Academics can work out techniques before they are ready for commercial investment. In-house efforts are often assembled on an as-needed basis, reflecting historical biases rather than optimal workflows.
Commercial software may be more optimized generically but is less well-suited to the idiosyncratic needs of individual pharmaceutical companies.
Among other issues, the interfaces between open-source software and other programs, or even interfaces with users, often require programming expertise beyond most so-called wet or bench scientists, or even bioinformaticists who have an expertise in programming but whose training is in bio-sciences.
Delano hopes his PyMOL can become a nucleus for advancing open-source efforts. All of the code is openly available; the company's income comes from service contracts. It's an unusual business model, he conceded. "When you're starting from an ideological standpoint, you've got to be constrained by that."
-Cost Molecular Modeling">
PyMOL models the three-dimensional structure of proteins. That can help scientists understand disease and design better drugs.
Many new drugs work by binding to proteins. By helping researchers see how a potential drug molecule nestles into the nooks and crannies of a specific protein molecule, PyMOL can help researchers imagine how to improve the drug or choose among different drug molecules based on how well each fits with proteins. According to the company's Web site, the software has been downloaded over 100,000 times.
Companies that download PyMOL are free to use it as a starting point to develop their own proprietary modifications. And though Delano dreams of communities of chemists and computer programmers contributing modules to make PyMOL more robust, too few have materialized. Other open-source products include Jmol, Chemistry Development Kit, Adaptive Poisson-Boltzmann Solver, Molecular Modelling Toolkit and GROMACS.
Plenty of code gets written at pharmaceutical companies, but they are reluctant to allow such work to be shared, said session organizer Andrew Smellie, senior investigator at cancer therapeutic company ArQule. The industry is "fanatic about IP [intellectual property] and it spills over not just to their [drug] compounds but to the rest of their business," he said.
That attitude could change as information technology becomes more integrated with drug discovery and development, hinted Rozwell, particularly because many biopharmaceutical companies feel they've made big investments in IT with little to show for it.
In addition, she said, some companies are trying to figure out how to use, rather than simply protect, intellectual property. "If you just have a patent, and you're not doing anything with it, it actually is not an asset anymore," she said. "So if you're managing your IP [intellectual property] well, you're a good candidate for [thinking about] open source."
Rozwell thinks something similar is already occurring when biopharmaceutical companies are highly involved in improving software vendors' products. The end result is similar to open source, she said. "It's a collaboration that will not only improve a product for the purchaser but that will benefit the whole industry."
An example of this was also presented at the conference. Daniel McMasters, senior research chemist at Merck, described adapting subscription-based software from Spotfire so that biologists could readily compare data across different types of assays, classes of chemical compounds and even individual test runs.
The Spotfire project began with frustrated biologists, who felt that existing techniques to summarize data hid crucial information about how cells responded to varying concentrations of different drugs.
"Every Friday, the biologists were looking at curves all day, highlighting data they didn't trust," recalled McMasters. So he and his team tweaked the program to reduce "artifacts of data reduction." For example, if separate assays measuring a drug's efficacy showed incompatible results, the program readily allowed researchers to drill down into individual test runs to look for anomalies.
Though the Spotfire staff participated in the project, it hadn't anticipated this particular use, said McMasters. Now, he said, it should be available to others.