these offer powerful targets for the understanding and manipulation of the system (Ideker et al., 2001). The central task of a systems approach is to (a) comprehensively gather information from each of the distinct levels, (b) examine relationships among the agents of the system, (c) hypothesize system topologies, (d) integrate data into predictive mathematical models of the system, (e) test predictions, and (f) identify key regulatory signals and relationships where intervention could stimulate new outcomes.

There are a growing number of publicly available molecular databases and systems analysis software programs that could be used for initiating systems modeling of social, behavioral, and genetic interactions. For instance, the Database of Interacting Proteins (Xenarios et al., 2001), the Biomolecular Interaction Network Database (Bader et al., 2001), and the Munich Information Center for Protein Sequences of the German National Center for Environment and Health (Mewes et al., 1999) contain searchable catalogs of known protein-protein interactions; the Transcription Factors Database (Wingender et al., 2000) and The Promoter Database of Saccharomyces cerevisiae (Zhu and Zhang, 1999) catalog interactions between proteins and DNA (i.e., transcription factor interactions), and databases of metabolic pathways also recently have been established (e.g., EcoCyc [Karp et al., 2000], KEGG [Ogata et al., 1999], and What Is There [Selkov et al., 1998]). A growing number of databases are also under development for storing the now sizeable number of mRNA-expression datasets (Ermolaeva et al., 1998; Stoeckert et al., 1999; Hawkins et al., 1999; Ringwald et al., 2000; Aach et al., 2000); companies, such as Affymetrix, Rosetta, Spotfire, Informax, Incyte, Gene Logic, and Silicon Genetics, market gene-expression databases commercially. Notably lacking from this list, however, are repositories of information on the behavioral and social components of the system. Work toward developing publicly available information on these levels could open up significant possibilities for the computer modeling of health outcomes.

The development and practice of systems approaches to model social, behavioral, and genetic interactions involves a number of requirements that will pose particular challenges for researchers. These include: (a) bridging disciplinary and language barriers encountered by teams of social scientists, behavioral scientists, molecular biologists, geneticists, and computational scientists; (b) the need for high-throughput facilities for molecular technologies, such as DNA sequencing, DNA arrays, genotyping, proteomics, metabonomics, and tissue arrays; (c) a lack of integrated public health, medical, and biological informatics systems; (d) the need to develop novel analytical tools and efficient, powerful computational infrastructures; (e) a lack of integration of discovery-driven and hypothesis-driven science; and (f) the need to develop diverse partnerships among academia, community, industry, and government.