Computational biology has officially arrived at the National Institutes of Health.
In an unprecedented event, the trans-NIH Biomedical Information Science and Technology Initiative (BISTI) gathered nearly 600 computational biology experts — along with NIH director Elias Zerhouni and the directors of numerous NIH institutes — on the NIH campus in Bethesda, Md., Nov. 6-7 for the first BISTI Symposium: “Digital Biology: the Emerging Paradigm.” The meeting capped off “Digital Biology Week” on the campus — five days of related events that, for many, signaled NIH’s long-overdue recognition of information science as a necessary component of biomedical research.
As one attendee noted following an all-star panel discussion, “If the NIH director and the heads of five institutes are on stage at the same time discussing computational biology, that’s pretty groundbreaking.”
The BISTI Symposium was a long time in coming. The consortium was first established in 1999, following the release of the so-called BISTI report, which was drafted by a working group of around 20 external advisors and laid out two principal recommendations: the creation of an NIH-wide body to oversee biomedical computing activities across the institutes; and the development of a network of national biomedical computing centers. The first goal was accomplished nearly immediately, with the establishment of the BISTI consortium, but its progress was stymied because a full-time BISTI chairman was not appointed until May 2003, when Eric Jakobsson was hired as director of the Center for Bioinformatics and Computational Biology at the National Institute of General Medical Sciences [BioInform 06-02-03].
Jakobsson immediately began working toward making the second goal a reality, and his efforts bore fruit with the release of the $2.1 billion NIH roadmap in September: As part of Zerhouni’s plan to “turbo-charge” NIH, $14 million to $17 million in FY 2004 funds were earmarked to create several national centers for biomedical computing (NCBCs). [BioInform 10-06-03].
The BISTI symposium was in part a kick-off party for that proposed infrastructure: In his introductory remarks, Jakobsson praised the NIH for its commitment to building a biomedical computing network, noting that the time may have finally arrived when the people who build such infrastructures are “valued equally with the people who do the discovery.”
Zerhouni remarked that even before he took the position of NIH director in May 2002, he thought that “of all the NIH initiatives, BISTI was the most important one to follow up on.” In planning the NIH roadmap, there was “consensus” across all the institutes that it was time to “accelerate what BISTI had proposed,” Zerhouni said, adding that “NIH understands that there are two critical components” to an effective biomedical computing infrastructure: widespread access to algorithms, tools, and computational power “to give scientists the capabilities to ask questions;” and a solid investment in “getting better data than we have today.”
The Road Ahead
Despite the warm-and-fuzzy sentiments from NIH leaders, the symposium wasn’t entirely a mutual admiration society. The agenda featured presentations by NIH Institutes and NIH-funded external researchers who were not only eager to share their biomedical computing success stories, but also happy to suggest specific problem areas where the NIH could focus its attention in the future. During afternoon concurrent sessions, conference attendees split into groups to discuss three primary biomedical computing challenges that NIH has identified: data integration, quantitative biology, and grid computing. In addition, the meeting’s keynote talks addressed a number of hurdles NIH is bound to face as it embraces digital biology.
In one keynote, Nobel laureate Sydney Brenner (who quipped that the term “digital biology” sounds like it refers to “the biology of the fingers”) stayed true to his reputation for stirring up controversy by denouncing current gene-based methods of reconstructing biological systems. “The gene is not the correct unit we should be looking at,” he said. “The correct unit should be the cell.” Introducing the Cell Map project he’s been promoting for at least a year [BioInform 09-16-02], Brenner advocated an approach that relies on complexes of interacting gene products rather than the gene products themselves. Brenner also criticized “top-down” approaches to modeling biological systems, noting that “people used to call that ‘physiology.’”
The main obstacle to future biomedical computing projects, according to Brenner, is the quality of the data available to researchers today — in particular microarray data, which he dismissed as “inaccurate and insufficient.” The goal for any data-generation project, he said, should be a data set that is “complete, accurate, and permanent.”
Brenner proposed that, just as molecular biology disappeared as a subdiscipline as it was absorbed into biological research as a whole, biomedical computing — whether it’s termed “digital biology,” “systems biology,” or “integrated biology” — will succeed “when everyone is a computational biologist.”
Providing a bit of perspective from the computational side of the equation, Microsoft CTO-turned-venture capitalist Nathan Myhrvold delivered another keynote that compared the current state of computational biology to “the state of computing in 1959.” While the big growth period in the field lies ahead, Myhrvold warned symposium attendees to pace themselves: “It isn’t a sprint, it’s a marathon,” he said.
Digital biology faces the same danger that many other technologies have already suffered from, Myhrvold said. Like artificial intelligence and, more recently, nanotechnology, Myhrvold said there’s a tendency for new approaches to be overhyped in the short run, and underhyped in the long run,” so it’s important to find equilibrium between raising public awareness and overpromising, he said.
In addition, Myhrvold cautioned attendees about the drive to set standards for biological data. The trick, he said, is finding “a balance between innovation and commonality. You don’t want to legislate out innovation” by enforcing strict data standards, he warned.
Genomics is just the “first example of a big data problem in biology,” Myhrvold said, but many more are on their way. Like Brenner, Myhrvold provided an example of how the biomedical computing community will know it’s been fully accepted by the scientific community: “When there is a Nobel prize awarded for a bioinformatics algorithm.”
One of BISTI’s goals in hosting the symposium was to provide opportunities for NIH program directors to brainstorm with the biomedical computing community on ways to help guide future funding decisions. A final report on the symposium’s activities will include recommendations from the three concurrent sessions that were held to address what NIH sees as “unmet technology needs” in biomedical computing: data integration, quantitative biology, and grid computing.
Greg Farber of the National Center for Research Resources, who moderated the session on data integration, told BioInform that the findings from the sessions might ultimately lead to new requirements for NIH funding. In the data integration session, for example, attendees reached consensus on eight recommendations that ranged from the creation of “standard” methods for representing biological data to appointing community advisory boards for large-scale community resources to ensure interoperability. Farber said he welcomes input from the broader bioinformatics community before finalizing the report. Those interested in receiving a draft can contact him directly at [email protected].
Reports on the entire symposium, as well as Digital Biology Week’s satellite events, will be released in the next few months. Any future changes in NIH’s biomedical computing strategy will likely draw upon those recommendations, just as its recent commitment to fund the NCBC network grew out of the original 1999 BISTI report. Of course, the digital biology community is hoping that the process is a bit quicker this time around, and, according to Jakobsson, it will be. Initial NCBC grant announcements are expected as early as this spring, he told BioInform. In addition, he said, NIH is also about to announce a companion program to NCBC that will support individual investigators who will work with the centers in an effort to support “both big science and small science.”