Post-Genomic Biology: Gene Expression Profiles, Cluster Analysis and Beyond

Abstract

multidimensional space in which each dimension represents a gene, and the position on the respective axis determines the level of expression. Thus, a single point in state space represents a distinct cellular state with respect to gene expression. REVIEWS g r a f t t x . c o m g r a f t j u l y / a u g u s t 2 0 0 1 v o l u m e 4 i s s u e 5 3 3 5 © 2001 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at LOCKSS on December 24, 2007 http://gft.sagepub.com Downloaded from philosophers of science would contend that being able to re-enact the full complexity of a real organism as a 1:1 model in the computer is not equal to a conceptual understanding of that complex system. The very existence of clusters of expression profiles that represent phenotypes such as cell fate, cell type, response to external perturbations or disease conditions points to a robust, self-organizing behavior that emerges from the collective action of the genes. The use of highly simplified models of the genetic network has been instrumental in providing insights into the fundamental, generic principles of how a large system of interacting genes gives rise to its “global observable,” the cell behavior with the constraints. The theoretical treatment of large model networks of anonymous genes suggests that the global dynamics of gene-gene interactions, as defined by the wiring diagram of the network, can be represented as a smooth landscape (in the abstract, high-dimensional gene expression space) which contains multiple attractors that act like energy minima. In that attractor landscape, a given gene expression profile at a given time, representing a cell state, is just a single point. Biological processes driven by the change of the expression profile can then be pictured as a marble (whose position represents the cell state) that is forced to roll along specified paths, the valleys, into attractor basins and end up in one of the attractors. The latter corresponds to a stable cellular phenotype, such as proliferation, differentiation or functionally activated state. The structure of the attractor landscape which is defined by the genomic wiring diagram determines how the gene expression profile changes; thus, it embodies the dynamic constraints and rule-like behavior of physiological and pathological cell processes, including self-stabilization, directionality and determination. In this model, environmental influences affect the position and course of the cell state by hitting the marble, which then has to roll in accordance with the attractor landscape sculpted by the genome. Thus, the model unites genetic determination with environmental inputs within one formal framework. Although these integrating concepts are still in their infancy, they will provide valuable guidance in exploiting the marvelous post-genomic technologies to acquire genome-wide data in a more systematic and purposeful way. This novel union of formal theory with high-throughput data acquisiREVIEWS 3 3 6 v o l u m e 4 i s s u e 5 j u l y / a u g u s t 2 0 0 1 g r a f t g r a f t t x . c o m Figure 5. Overview of the successively increasing level of integration in the interpretation of gene expression profiles (left) and the information gained (right).