HINXTON, UK--In the year since reusable and collaborative components were called "hype" at Objects in Bioinformatics `97, components have become reality and their standardization has matured rapidly.
Organized by Alan Robinson of the European Bioinformatics Institute (EBI) outstation of the European Molecular Biology Laboratory, the Objects in Bio in formatics conference held at the Wellcome Trust Ge nome Campus here August 3-4 attracted more than 220 participants from academia and industry.
Presentations featured discussions of the implementation and application of distributed object technology such as CORBA to genomics and bioinformatics, and the work that the Life Sciences Research Task Force of the Object Manage ment Group has done to promote standards.
Jon Siegel, the group's director of program management, described how interoperability will be facilitated through object-oriented standards. Based on CORBA, the Object Management Group offers a framework through which various domains can propose and implement standardized interfaces and services to domain-specific data and components.
Dave McAllister of Silicon Graphics offered a slightly different view on components. Java and JavaBeans, he contended, are not the answer to a component model and are only to be used for lightweight clients based upon such a model. Other programming languages are better suited for heavyweight applications, he said.
McAllister also argued that software development tools are still prehistoric and not suitable for today's distributed applications. Such tools need to become object-oriented themselves. The lack of proper and modern software development tools was echoed in many of the talks that followed.
Good uses for reusable components
Other talks demonstrated the good use to which reusable components and the CORBA architecture for data exchange have been put. For example, Jeremy Parsons of EBI presented ongoing work that is aimed to make tools to explore EST data sources, where CORBA is used to provide an interface between distributed data and the Java applications that view sequence chromatograms, alignments, and clusters.
Another typical example was presented by Richard Baldock, MRC Human Genetics Unit, who introduced the audience to work on a digital voxel 3D-image atlas of mouse development and a gene expression database that uses the atlas as a reference. The underlying database of embryonic objects (tissues, reconstructions, and nomenclature) is implemented in ObjectStore, using CORBA to connect the database to Java applets that will, in the near future, enable researchers to query the database and even to submit new gene expression data.
Although reusable components for bioinformatics are what this conference was all about, a lot of individual solutions to certain problems were presented. Stan Letovsky of Cereon Genomics commented, "Everybody is building reus able components but everyone is building the same reusable components over and over again. It is getting to the point where there is not a whole lot of reason for somebody to build yet another sequence viewer, yet another alignment viewer."
Ira Baron of Millennium Pharmaceuticals commented that another reason for multiple implementations of the same sort of tool "may be simply because many of the participants here are commercial research houses or software vendors so there may be copyright reasons for not being able to use somebody else's implementation."
CORBA is the glue
CORBA is now playing an even more important role than last year in gluing applications together; many of the presenters had adopted this architecture over any other. Baron said he was impressed: "I don't think there are many industries where CORBA has come into common use as quickly as in the bioinformatics community. Some of that might be a reflection that there are a lot of legacy systems in bioinformatics relative to other industries, and some of that might also be that the nature of the science in bioinformatics is very distributed, much more so than in banking or insurance, for instance."
The general consensus here was favorable for reusable components. "A year ago we were talking of these things and we had some examples but it was not as mature as it is now. We now have three or four groups putting out sets of new components," Letovsky said. The next phase will be to play, and play with components from different authors, he concluded.
During the panel discussion, a proposal was brought forward to soon have a standard interface for selecting objects, to make a start with the plug-and-play components of different authors.
With real interoperability and reuse of these bioinformatics components just around the corner, it was no surprise that many voiced the same opinion as the Sanger Centre's Ewan Birney: "Modular components are the only way to go forward in bioinformatics."
--Jean-Jack M. Riethoven