Objects in BioInformatics '97: the report on the conference and an interview with the organiser
Report of the conference
Interview with the organiser
Java and CORBA are good, reusable and collaborative components are the hype, and standardisation is the key issue. These are in short the conclusions of the Objects in Bioinformatics '97 conference (OiB'97) that was held in the new Conference Centre at the Wellcome Trust Genome Campus (Hinxton, United Kingdom) on 18 and 19 June.
The conference was jointly organised by the European Bioinformatics Institute (EMBL-EBI) and the bioWidget Consortium, and focused on the growing role of reusable object oriented software components and distributed objects in bioinformatics and the biological sciences. "The widget and object-oriented approach is a powerful paradigm in which new applications are developed by utilising pre-existing components interacting in concert, and offers the possibility of a new generation of bioinformatic tools. In this way development time is reduced and greater time is spent on data analysis and knowledge discovery, rather than on the writing of monolithic applications", Alan Robinson of EMBL-EBI explains, the organiser of OiB'97. "Initially, we were aiming at a small workshop of 40 to 50 persons, but we were overwhelmed at the number of people registering. Approximately 200 official registered scientists, software developers, and managers attended the conference, with another 50 or so unregistered visitors."
BioWidgets and others like that
A sizeable part of the first day was dedicated to the bioWidgets Consortium. Stan Letovsky, of the John Hopkins University and member of the Board of Directors of the bioWidget Consortium, described the Consortium as a non-profit organisation with the function to "find better and more economical ways to develop graphical displays for genomic data, and biological data in general. It does this by developing a community consensus on standards for graphical components (bioWidgets), to implement bioWidgets, and to build genomic graphical applications from the bioWidgets." It has been decided that Java will be the language for bioWIdgets, as it is a standard language, encourages modular and cross-platform development, and is seen by the majority as cutting edge technology.
Further talks discussed the bioWidgets low-level Application Programming Interface (API) and the Quality Assurance (QA) process that every widget has to go through before it gets the label 'bioWidget'. It was announced that the bioWidgets Consortium will probably extend its work towards objects for standardised database access, since these methods for retrieving data are often tightly coupled with the widgets that display the actual data.
The Conference Hall: lecture on the bioWidget API.
(Copyright 1997 The BioInformer)
Developers' paradise: boot camp for bioWidgets
A boot camp will be organised in Bar Harbor at the Jackson Laboratory on 26 October to 11 November 1997. The target audience are active developers of bioWidgets code; the purpose of this camp being to bring people together in a working environment to share ideas and code, and to make progress on common architecture. Integration of several bioWidgets is also planned during that week.
David Benton, of SmithKline Beecham Pharmaceuticals (Philadelphia, PA, USA), outlined a new proposal (called 'BioMolecules') to extend the model of the bioWidgets Consortium to other areas in bioinformatics, e.g. algorithmic engines for common software tasks like sequence similarity searching or multiple sequence alignment, and access to local and remote databases. He proposed to utilise Common Object Request Broker Architecture (CORBA) for this effort, and to affiliate with the Object Management Group (OMG).
API for bioWidgets will be released soon
One of the major results of this part of the conference "is the bioWidgets API, which is going to come out within the next couple of weeks, and will probably be published in CABIOS or a similar outlet", concluded Suzanna Lewis of the University of California and member of the Board of Directors of the bioWidget Consortium. She remarked on the doubts some of the attendees had about the usefulness of the OMG BioMolecules proposal: "I absolutely think the bioWidget Consortium and the OMG BioMolecules can coexist, because the two are very separate; where the bioWidgets focus on graphical views, the OMG BioMolecules is about the datamodels, and it is the datamodel that you pass to a (bio)widget. The two are tightly coupled. I think this a very intriguing proposal and I intend to be involved myself."
Since visualisation of genomic data is considered very important, Hank Schiffmann of Silicon Graphics Inc. (SGI), provided an energetic talk about approaches how to create 3D interactive (bioinformatics) applications on the web. Java alone is too poor in 2D and 3D graphics to be any good on its own, and "ActiveX is the worst possible solution". Three better solutions were discussed: Java bindings to OpenGL, the SceneGraph API that SGI has been working on, and VRML. The first alternative is the most flexible and programmable, but also costs the most effort. With the SceneGraph API applications are easier to write and have better performance, while VRML was pictured as the graphic artist solution where minimal code will be used. The proposal is to add an external authoring interface to VRML, which basically will add the SceneGraph API to VRML.
During the remainder of the conference, the focus went to case-studies of the object-oriented approach in bioinformatic tools, and to recent developments in distributed, collaborative, and interoperable objects.
More about those objects
Analysing the presentations of the speakers, the posters, and the demonstrations, there is a definite push towards reusable objects and especially interoperable or collaborative objects. Let's clarify the differences between all the objects in the next descriptions: Reusable objects are software components that have been carefully designed, with a well specified intuitive interface, so that they can be used in future, possibly unprecendented, environments. Objects that can invoke operations on one another through a common communication protocol (be it IIOP, sockets or whatever) and a common, well-defined programming interface are generally called interoperable objects. Collaborative objects are more closely coupled - they interact more dynamically to query and change each others properties and state. Distributed objects are those that are located in diverse computational environments (e.g. on different types of machines in different countries). Technologies like CORBA facilitate interoperability and collaboration between distributed objects.
The Cloisters the day before the conference...
(Copyright 1997 The BioInformer)
... and during the tea break and poster viewing sessions.
(Copyright 1997 The BioInformer)
Although the first part of the conference was dedicated to the bioWidgets Consortium, other initiatives towards reusable componentry in bioinformatics were also presented. Stephen Chervitz of Stanford University (California) discussed his group's project 'Bioperl' in which they aim to develop a set of object-oriented Perl modules to manipulate molecular information. The Bioinformatics Template Library (BTL), focused at providing a library of C++ algorithms and data structures for bioinformatics and molecular modelling, was outlined by David Moss of the Department of Crystallography, Birckbeck College (London).
Java, however, was one of the most prevalent languages of choice in this arena, but "there seems to be a general consensus, a convergence to adopting CORBA as a way of data exchange", Suzanna Lewis concluded.
Legal consensus is required
David J. Parry-Smith, of the Bioinformatics Group at Pfizer Central Research (UK), said about the different initiatives in creating bioinformatic objects: "I think there has been a lot about widgets, interoperable objects and developing software collaboratively, reusing software, all those things that we all knew before we came here. But I think what's come out of it is that there is a strong need for standardisation somewhere along the line, not everyone is clear about where that standardisation line should be. One of the issues that underlies this all is the legal side of how to actually share code. The bioWidget Consortium hasn't been able to come to an agreement about licensing yet, and other initiatives, like the Open Molecule Foundation (OMF), have had similar discussions and haven't come up with an acceptable agreement either. So without some move on that front it is going to be very difficult for us to share code, so some of these ideals of code reuse come actually to perish."
Suzanna Lewis countered "in my opinion it is possible to share code and I really think that whether you are commercial, academic or non-profit that you'll gain from sharing code and ideas and using these for a bigger creative outlet in whatever it is you're doing."
Her sentiment was shared by many of the presenters and attendees who spoke out the desire to cooperate with the bioWidget Consortium, or with one of the other complementary initiatives. It seems that for academic and non-profit organisations the legal issues involved are only secondary to making sure they are not left behind in this new boost towards the development of a new generation of bioinformatic applications.
A week or two later....
After the dust of the conference settled, The BioInformer met with the organiser Alan Robinson for a look back:
On the lighter side: Alan's room the evening before the conference.
(Copyright 1997 The BioInformer)
Considering the large number of people attending (versus your initial 40-50 estimate), do you think re-usable componentry in bioinformatics is an issue that everybody (both scientists and managers) is aware of and supports?
From the response of the audience of OiB-97, and certainly at the research level, the concept of libraries of high quality, freely available, reusable components that can aid in research is an extremely attractive proposition. I feel that this is a issue about which higher management is probably becoming aware because of the realisation that much time is being wasted on reinventing the wheel when scientists should be concentrating on what they are best at, which is research. The widget and object-oriented approach is a powerful paradigm in which development time is reduced and greater time can be spent on analysing the data by developing new applications from existing components. That, I believe, is the message that should be getting back up the management ladder.
Looking back over the two days, what do you think are the conclusions from this conference?
I came away believing that there is a great deal of interest in projects such as the bioWidget Consortium. Within Europe, there has already been a tremendous amount of development using Java and investigation of how CORBA could be used to share information. OiB-97 was an opportunity for people from both the US and Europe to see what was being done and to establish contacts. People not only want libraries of good quality components available so that they do not have to write yet another sequence viewer, but they want also to contribute their own widgets and become involved. The community realises the strength and benefit of these endeavours to itself. We are in a situation where the bioinformatics community could become proactive and produce, validate and release much software that it requires and which all works in unison.
We have seen several proposals/actions with respect to building reusable components. Overkill or the more the better?
There are other actions whose goal is to develop reusable components and all share a similar philosophy. The main difference between the actions is usually the language in which the components are being developed. The bioWidget Consortium has chosen Java, as has the OMF, although the OMF has a different mandate to the bioWidget Consortium. Although the languages may differ, I believe the projects share much common ground and similar obstacles to overcome and for this reason, I believe, dialogue between the organisations will lead to better solutions and implementations.
The general consensus seems to be directed towards Java (and possibly CORBA) as the main developing language/layer. Are the other object building actions based on Perl/C/others doomed to fail from the start?
There will always be many programming languages and each will have its own strengths as well as weaknesses. Java has a high public profile and it is a language that many people have now learnt to program in. Java's strengths include its platform independence, its object-oriented nature and its security model. Right now, I would say these features make Java the best language for the development of a library of reusable widgets to be used in a heterogeneous environment. However, Java is not good at dealing with strings or parsing flat files, but these are obvious strengths of Perl. For the time being, I foresee both the BioPerl and bioWidget efforts necessarily co-existing. With time- and mission-critical applications, C++ will remain the industry-strength choice for the time-being, although there is a clear swing towards Java. With the uptake of technologies such as CORBA, these issues become more academic since with a clearly defined interface, objected-oriented classes, whether they be written in Java, Perl5, Smalltalk or C++, will be able to interoperate and communicate in both local and distributed environments.
Any issues that have not been resolved during this conference?
The main issues that were not resolved adequately include the legal discussions about the licensing of the software, and the release of an API to which people may write. The legal issues may prove to be the most problematic, especially with regard to those contributing software from a non-academic position. A new version of the API will be released shortly and made available from the bioWidgets home page.
This is the first conference in the Genome Conference Centre. What has been the response to the facilities from the participants?
Any plans in organising another conference in the near future?
Organisation has started on a conference to be held on the 10th and 11th of December 1997 at the Wellcome Trust Genome Campus. The conference, which is the idea of Paolo Zanella (Head of the EBI) and Christopher Jones (on sabbatical from CERN to the EBI), is entitled Frontiers in Visualisation and Computer Human Interaction. It will aim to expose not only researchers and managers of the bioinformatics community to visualisation techniques and better ways to interface with computers, but also to educate the computer graphics industry to the problems of biology and bioinformatics. Already, there are a couple of world-class speakers organised and a Web page describing the conference is available.
Other remarks you wish to make?
Given the success of this year, I feel it is inevitable that OiB-98 will happen. Already I have had people asking if it is being planned, and have had offers of sponsorship. I also believe that we will once again hold it at the Wellcome Trust Genome Campus, by which time the on-site accommodation will have been finished, as well as an extra bar!
More pictures of the conference on a separate page. Note: approximate loading time: 3 minutes (14k4 modem), 80 seconds (33k6 modem), 20 seconds or less (ISDN and better).
Article by: Jean-Jack Riethoven
Resources and further information
European Bioinformatics Institute
Frontiers in Visualisation and Computer-Human Interaction conference
Objects in Bioinformatics '97
Alan Robinson's homepage
Info about the developer's boot camp:
Contact Matthew Cowley (email@example.com)
Object Management Group
About 'BioMolecules' proposal
e-mail: David Benton(firstname.lastname@example.org)
BTL Bioinformatics Template Library
External sites are not endorsed by EMBL-EBI