SALT LAKE CITY, Utah--The seven-month-old quest for bioinformatics software standards reached a significant milestone here February 13, when a working group drafted the first formal request for proposals from software vendors interested in creating standard software objects for analyzing genetic sequences. If approved at a March 30 meeting in Manchester, UK, the request for proposals could produce by late 1998 the first widely accepted rulebook for developing certain kinds of interoperable bioinformatics tools and services.
In other action, the standards-seeking work group, the Life Sciences Research Domain Special Interest Group, extended to March 9 the deadline for corporate and academic researchers to provide comments on what form common standards for an Interface Definition Language (IDL) should take. So far, the group has received comments from 18 companies and academic institutions. The comments, and other information about the group's activities, are posted at http://www.omg.org/lsr/.
"Things look like they're moving quickly, and interest continues to grow," Eric Neumann of Netgenics, a Cleveland-based software company, told BioInform. He chairs the Life Sciences group along with David Benton of Philadelphia-based SmithKline Beecham and Tim Slidel of the European Bioinformatics Institute (EBI) in Hinxton, UK. The group operates under the auspices of the Object Management Group (OMG), a nonprofit consortium based in Framingham, Mass., that promotes object-software standards based on its Common Object Request Broker Architecture (CORBA).
The effort to establish IDL standards--begun at a Philadelphia meeting of more than 50 bioinformatics professionals in August, 1997, (see BioInform, September 1, 1997) is intended to eliminate the many obstacles that currently prevent informatics researchers from easily sharing data and software. Academic researchers have long been frustrated by the problem, but solving it became a priority when major pharmaceutical companies realized they risked investing millions of dollars in bioinformatics systems that might not communicate.
Consensus-Building and Priority-Setting
To forge a consensus on standards, software vendors and users leapt into the OMG's highly structured decision-making process (see http://www.omg.org ), which brings interested parties together in a variety of working groups. The groups produce proposals for standards that are voted on by OMG's 800 members, including most of the computer industry's leading companies and other major corporations and universities. According to OMG, both technical and business considerations guide the acceptance of standards.
Neumann said his group is "optimistic" that the draft Request For Proposals on Sequences and their Analyses prepared here will be approved at the Manchester meeting. Once approved, "interested vendors would have 140 days--until the end of August--to submit a sequence analysis proposal, complete with IDL specifications," he explained. "This will signify [the group's] first attempt at producing a standard."
The Life Sciences group has also drafted "Roadmap" and "Architec ture" documents, Neumann reported, which will be used to set priorities for the preparation of requests for information and proposals from the many life sciences disciplines interested in setting IDL standards. One group interested in cheminformatics, for instance, will meet in Manchester to discuss its needs. A Frameworks Working Group is also gathering information, partly from the 18 sets of comments the group has received, on how to define broader frameworks for interoperability within the life sciences community.
Corporate and Academic Comments
The comments, including submissions by companies such as Netgenics, the La Jolla-based Sequana Therapeutics division of AxyS Pharmaceuticals, Oxford, UK-based Oxford Molecular Group, and academic groups such as the EBI, the University of Pennsylvania's Bioinformatics Center, and the collaborative "Bioperl" software-writing group, reflected diverse interests, from genomics and genetics to computational chemistry and computational molecular biology. Most of the comments highlighted ongoing software development efforts that could play a role in devising the new standards, but a few offered broader comments on how to achieve the long-elusive goal of interoperability.
Leading bioinformatics software vendor Oxford Molecular, for instance, concluded that "the users of bioinformatics software are better placed to judge what changes will improve its quality and utility than are the software vendors. The mission of the group is thus more likely to be achieved if the focus of the…standardization initiative is determined principally by the users of bioinformatics software, rather than by the vendors." The company noted that its submission was designed more to "support discussion within the group…than to attempt to influence the direction of the standards initiative."
For their part, the researchers involved in the Bioperl effort--including Stephen Chervitz and Steven Brenner of Stanford Univer sity in Stanford, Calif.; Georg Fuellen of the Universitaet Biele feld in Germany; Chris Dagdigian of the Genetics Institute in Cam bridge, Mass.; and Richard Resnick of Williamstone Enter prises in Waltham, Mass.--were "eager to join the effort," in part because they believe the adoption of CORBA-based standards would support their view that "the design of bio-objects should be language-neutral."
"It is most expedient to focus on the best features of different languages rather than to try and use a single language for everything which does only a few things optimally," the Bioperl group wrote. "Certain functionalities are more easily and efficiently implemented in certain languages. For example, Perl's text processing and regular expression support make it well-suited for complex text parsing tasks (common in bioinformatics)….A technology like CORBA can enable this strategy of 'the right task for the right language.'"
In its comments, Cherwell Scientific of Oxford, UK, maker of the Java-based software product ChemSymphony, argued, "the scope for inventing specific standards in the fields of cheminformatics and bioinformatics is narrow and should be kept narrow! It is arguable that there has been too much invention and proliferation of proprietary or case-specific standards by commercial and academic research organizations."
Cherwell said, "it is in the area of communication between applications and domains and the exchange of complex data that standards will be most helpful." The company also suggested that "it is important--but it is also simple and naked self-interest--that commercial or noncommercial software developers who produce object-oriented solutions should publish their interfaces.…Unless software developers do this it is more difficult for their solutions to interoperate with whatever else the scientific community may need to deploy. From this standpoint we believe that one useful achievement of the group would be to indicate that some objects, widgets, beans, or applets are deemed to meet the basic requirements of whatever standards are agreed."
Whether all--or any--of this blizzard of advice will be reflected in the final standards remains to be seen.