The Genom of Homo sapiens.pdf

Systems Approaches Applied to the Study ofSaccharomyces cerevisiae and Halobacterium sp.A.D. WESTON,* N.S. BALIGA,* R. BONNEAU,* AND L. HOODInstitute for Systems Biology, Seattle, Washington 98103-8904Integrative systems approaches to studying biologicalsystems have begun to yield striking results. Systems biologyhas emerged as a powerful new approach over thepast 5 years because of (1) the completion of the humanand many other genome sequences, which led to the identificationand prediction of comprehensive gene lists; (2)the development of high-throughput techniques for genomics,proteomics, metabolomics, and phenomics, leadingto the acquisition of global data sets; and (3) the creationof powerful computational methods for storing andassessing different types of global data sets as well as analyzingand integrating them. The integration of differentdata types is essential because it helps to deal with noisethat is inherent in large data sets and it reveals new biologicalphenomena that are not obvious from the analysisof single data types. Here we summarize the initial applicationsof systems biology to two systems: one, whichhas been studied intensively for years (the galactose utilizationsystem in yeast), and a second (the oxygen andlight responses in Halobacterium sp.) for which little dataare available. Systems approaches have allowed us togain new and fundamental insights into both systems(Ideker et al. 2001; Baliga et al. 2002).A key aspect of systems biology is viewing biology asan informational science. First, there are two generaltypes of biological information: the digital information ofthe genome, and environmental information that interactsdirectly or indirectly with the digital genomic information.Second, the genome has two major types of digitalinformation: the genes that encode protein and RNAmolecular machines, and the transcription factor-bindingsites in cis control regions of genes that create the linkagerelationships for gene regulatory networks which controlthe temporal and spatial parameters of gene expression aswell as their amplitude. Proteins may act alone, in loosefunctional relationships or biomodules (e.g., the enzymesof sugar metabolism), in complex protein machines (e.g.,the ribosome), or in larger protein networks. Transcriptionfactors, co-transcription factors, and the factors thatmediate changes in chromatin structure all operate viaDNA-binding sites in cis control regions of genes to regulategene expression. The protein and gene regulatorynetworks, although conceptually distinct, obviously arefunctionally integrated with one another. Third, it shouldbe stressed that biological information is of many differenttypes, starting with the core DNA genomic informationand moving out to ecologies (DNA → RNA → protein→ protein structures and biomodules → networks ofbiomodules → cells → organs [tissues] → individual organisms→ populations of individual organisms → ecologies),and that environmental signals increasingly modifythe basic digital information as one moves outward in thisinformation hierarchy. Therefore, to understand systems,one must have the tools for global measurements of asmany information types as possible and the ability to integratethese different types of information. Finally, biologicalinformation operates across three distinct time dimensions:evolution, development, and physiology.Accordingly, the patterns of genomic content and organizationchange across evolution as the patterns of gene expressionchange across development or throughout aphysiological response. Indeed, the informational contentof cells or organisms may be viewed as a series of snapshotsof changing patterns of information expression.The systems approach may generally be described asfollows:• A biological system is chosen and all preexisting relevantinformation is integrated into a model that may bedescriptive, graphical, or mathematical.• A global analysis of the systems elements is carriedout. Generally this begins with a genome sequence.Genes (and their corresponding proteins) and transcriptionfactor-binding sites may be cataloged, predictedcomputationally, or experimentally identified.These data sets are the initial building blocks for constructingprotein and gene regulatory networks.• The system is then perturbed genetically or environmentally,and global data sets are collected from asmany different data types as possible. Genetic perturbationsinclude overexpression, underexpression, orknockouts. These data sets are generally collected understeady-state conditions. Environmental perturbationsmay include, for example, the introduction ofsubstrates that activate metabolic pathways, and hormonesthat trigger signal transduction pathways. Here,kinetic experiments are required so that the patterns ofinformation change across development or physiologicaltriggering can be characterized.• The different data types must be integrated (see examplesbelow) and then compared against the initialmodel. Discrepancies between data and model willemerge. Hypothesis-driven formulations must explain*These authors contributed equally to this paper.Cold Spring Harbor Symposia on Quantitative Biology, Volume LXVIII. © 2003 Cold Spring Harbor Laboratory Press 0-87969-709-1/04. 345