BALTIMORE--Discussions that described the power of comparative genomics, the necessity for annotation standards, and the importance of analyzing protein domains to determine function were highlights of the Third Annual Conference on Computational Genomics. Overall, presentations indicated that scientists who have approached genomic analysis from different disciplines are "starting to converge," organizers and attendees told BioInform.
The Institute for Genomic Research held the meeting, which drew 316 attendees, here November 18-21. According to Tony Kerlavage of Celera Genomics, who cochaired the event with Steven Salzberg of TIGR and David Searls of SmithKline Beecham, organizers are positioning the event as one of the three most important gatherings for bioinformatics professionals. (Kerlavage said he considered the International Conference on Intelligent Systems for Molecular Biology, which is held annually in August, and RECOMB, the International Conference on Computational Molecular Biology, held in April, to be other important meetings.) A five-person organizing committee was a key to the meeting's success this year, he added. Committee members were Peer Bork of the European Molecular Biology Laboratory, David Haussler of University of California, Santa Cruz, Gene Myers of Celera, Temple Smith of Boston University, and David States of Washington University, St. Louis.
Eight exhibitors at Computational Genomics were Data Unlimited, EraGen Biosciences, InforMax, LION Bioscience, Silicon Genetics, Spotfire, US National Center for Biotechnology Information, and Viaken Systems. Meeting sponsors were Compaq, DuPont, SmithKline Beecham, and the US Department of Energy.
Temple Smith's lecture, "Protein Domain Dissection and Functional Identification," was considered by some to be the meeting highlight. Smith stressed that sequence domains, which he described as evolutionary modules or LEGO bricks, are the keys to determining biochemical function. He discussed the use of various bioinformatics tools for identifying protein domains, but concluded that the scientific community's lack of consensus regarding those tools is a handicap. He proposed organizing a summer study session at which the world's top bioinformaticists could converge to determine a best-of-breed algorithm.
A lecture by David Eisenberg of the University of California, Los Angeles, titled "Protein Functions from Genome Sequences," was also widely considered to be among the meeting's most significant. Eisenberg, who said his lab works on some 30 fully sequenced genomes, discussed new computational methods that assign protein functions and interactions to those sequences. He stressed the value of public genomic data that groups such as the Institute for Genomic Research generate. "I feel like going to Rockville and falling on my knees and kissing the ground," he exclaimed. Eisenberg said the way the scientific community defines protein functions has changed. "Now we define a protein within the context of its interactions in the cell, the way we define ourselves by our own interactions within our environment," he remarked.
Attendees noted that Peer Bork's presentation, "Comparative Sequence Analysis: From Polymorphism to Protein Domain Annotation," illustrated the power of comparative genomics. John Moult of the University of Maryland presented analysis linking single-nucleotide polymorphisms to specific diseases. Andrea Califano of IBM's Computational Biology Center presented a new automated pattern- discovery approach that attendee Jean-Francois Tomb, senior research associate at DuPont, described as a "conceptually simple and elegant algorithm that is a clear departure from traditional methods of looking at sequence similarity." For a computational person in the audience, Tomb added, some of the meeting's discussions were like light bulbs switched on to reveal new opportunities for their work.