CULTURE AND RESEARCH INFRASTRUCTURE 353
- Potential impact. All else being equal, institutions would prefer to support research in which the
potential impact of success is large. However, as a rule, claims of large impact are much more specula-
tive than other claims, simply because the long-term ramifications of any given discovery are difficult to
underscore in any convincing manner before the fact. - Technical risk. A research investigation may or may not be successful. Research that presents the
lowest technical risk (i.e., the lowest risk of failure or of being unsuccessful) is most often very closely
tied to some existing and successful research. Thus, as a rule, research that is of low technical risk tends
also to be of lesser potential impact. - Personnel risk. Research is performed by people, and any given research effort can be executed
more or less effectively depending on the people involved. Established track records of success are an
important dimension of the teams proposed to undertake research but cannot be the only dimension
taken into account if new researchers with good ideas are to be welcomed. - Budget. Institutions with a fixed level of support to offer investigators can support a larger
number of inexpensive research proposals or a smaller number of more expensive ones. All else being
equal, inexpensive proposals will tend to be favored over expensive ones.
Proposals for research must weigh each of these factors and make trade-offs among them. For
example, a lower budget may mean greater technical or personnel risk; a high-impact project may have
greater technical risk. Funding agencies must assess the plausibility of the trade-offs that a prospective
research team has made.
These notions suggest that review panels need a wide range of expertise and experience to judge the
merits of new proposals effectively or to carry out peer review of scientific papers. In principle, the
requisite range of expertise can be obtained through the use of a set of individual disciplinary experts
whose collective expertise is adequately broad. An alternative is to use a few individuals who them-
selves have interdisciplinary expertise. The disadvantage of the first model is that for practical purposes
it may reproduce forums in which the difficulties of cross-disciplinary understanding are manifested.
The disadvantage of the second model is that such individuals may be few in number and thus difficult
to enlist.
10.2.5.2 Federal Support
A variety of federal agencies support work at the BioComp interface, and this support has grown
over time.
10.2.5.2.1 The National Institutes of Health For computational biology (i.e., the computing-to-biology
side of the BioComp interface), the main actor in the U.S. government is the National Institutes of
Health, part of the Department of Health and Human Services.
A notable instance of bioinformatics work at NIH is the National Center for Biotechnology Informa-
tion (NCBI), a part of the National Library of Medicine. Established in 1988, it is NCBI that created and
maintains GenBank (see Chapter 3).
The NIH’s National Institute of General Medical Sciences (NIGMS) manages the Biomedical Infor-
mation Science and Technology Initiative, or BISTI. BISTI represents an NIH-wide collaboration and
coordination program between its many institutes and centers, as computational biology and
bioinformatics activity is spread throughout the organization. In addition, NIGMS also runs the Center
for Bioinformatics and Computational Biology, which focuses on theoretical and methodological infra-
structure, such as modeling, simulation, theory, and analysis tools in biological networks.^37 The NIH’s
Center for Information Technology, in addition to providing IT services to the rest of NIH, also main-