368 CATALYZING INQUIRY
it, characterize it, understand it better, or make it more generally applicable to other problems. Thus, the
biologist will likely be interested in the results of a model run on the single dataset of interest, while the
computer scientist will want to run hundreds or thousands of datasets to better analyze the behavior of
the model, and mathematicians will want to explore the limits of a model’s applicability.^70
An example of this cultural difference is illustrated in the history of the Gene Ontology (GO) discussed
in Chapter 4. Begun in 1998 as a collaboration between researchers responsible for three model organism
databases (FlyBase [Drosophila], the Saccharomyces Genome Database, and the Mouse Genome Database),
GO collaborators sought to develop structured, controlled vocabularies that describe gene products in terms
of their associated biological processes, cellular components, and molecular functions in a species-indepen-
dent manner. In their work, these researchers have apparently not made extensive use of the (mostly
domain-independent) theoretical contributions of computer science from the last 20 years, but rather have
reinvented much of that work on their own. The reason for this reinvention, offered by one knowledgeable
observer, is that they were unable to find computer scientists with appropriately specialized experience who
were willing to sacrifice their quest for general applicability to develop a functional, usable system.^71
A related point is that in academia, research computer scientists have very little motivation to take a
software implementation beyond the prototype stage. That is, they may have developed a powerful
algorithm that is likely to be useful in many biological contexts, implemented a prototype software system
based on this algorithm, and convincingly demonstrated its utility in a few cases. But because most of the
intellectual credit inheres in the prototype (e.g., papers for publication and promotions), research com-
puter scientists have little motivation to move from the prototype system, which can generally be used
only by those familiar with the quirks of its operation, to a more robust system that can be used by the
broader community at large. Because going from prototype to broadly usable system is generally a time-
intensive process, many powerful methods are not available to the biology community.
Similar considerations apply in the biology community with respect to data curation. Intellectual
credit for academic biologists inheres in the publication of primary data, rather than in any long-term
follow-up to ensure that the data are useful to the broader community. (Indeed, if the data are not made
useful to the broader community, the researcher originally responsible for the data gains the competi-
tive advantage of being the only one, or one of a few, able to use them.) This suggests that cultural
incentives for data curation (or the lack thereof) have to be altered if data curation is to become a more
significant activity in the research community.^72
(^70) These differences in perspective are also found at the interface of medical informatics and bioinformatics. For example,
Altman notes that “the pursuit of bioinformatics and clinical informatics together is not without some difficulties. Practitioners in
clinical medicine and basic science do not instantly understand the distinction between the scientific goals of their domains and
the transferability of methodologies across the two domains. They sometimes question whether informatics investigators are really
devoted to the solution of scientific problems or are simply enamored of computational methodologies of unclear significance [emphasis
added].” To reduce these tensions, Altman argues—similarly to the argument presented in this report—that “informatics inves-
tigators (and their students) be able to work collaboratively with physicians and scientists in a manner that makes it clear that the
creation of excellent, well-validated methods for solving problems in these domains is the paramount goal.” See R.B. Altman,
“The Interactions Between Clinical Informatics and Bioinformatics: A Case Study,” Journal of the American Medical Informatics
Association 7(5):439-443, 2000.
(^71) Russ B. Altman, Stanford University, personal communication, December 16, 2003.
(^72) One approach that has been used to support data annotation and curation activities is the data jamboree. In November 1999,
the Celera Corporation hosted an invitation-only event (“the jamboree”) in which participants worked for two weeks at annotat-
ing and correcting data from the Drosophila melanogaster genome. By all accounts a successful event that resulted in the publica-
tion of the complete sequence as well as appropriate annotations (see M.D. Adams, S.E. Celniker, R.A. Holt, C.A. Evans, J.D.
Gocayne, P.G. Amanatides, S.E. Scherer, et al., “The Genome Sequence of Drosophila melanogaster,” Science 287(5461):2185-2195,
2000) the event featured a very informal atmosphere that promoted social connection and interaction as well as a work environ-
ment conducive to the task. The emergence of some level of community curation on Amazon and eBay may also provide some
useful hints on how to proceed. In these efforts, community assessment is allowed, but there’s no overall review of the quality of
the assessment. Nonetheless, users have access to a diverse collection of assessments of and can do their own meta-quality
control by deciding which of the reviewers to believe. This model does scale with increasing database size, although consistent
curation is hardly guaranteed. It is an open question worth some investigation as to whether community commentary (perhaps
supported with an appropriate technological infrastructure) could result in meaningful data curation.