consist of text strings indicating appropriate bases, but when there are gaps in sequence data, gap lengths (or bounds on gap lengths) must be specified as well.

  • Graphs. Biological data indicating relationships can be captured as graphs, as in the cases of pathway data (e.g., metabolic pathways, signaling pathways, gene regulatory networks), genetic maps, and structured taxonomies. Even laboratory processes can be represented as workflow process model graphs and can be used to support formal representation for use in laboratory information management systems.

  • High-dimensional data. Because systems biology is highly dependent on comparing the behavior of various biological units, data points that might be associated with the behavior of an individual unit must be collected for thousands or tens of thousands of comparable units. For example, gene expression experiments can compare expression profiles of tens of thousands of genes, and since researchers are interested in how expression profiles vary as a function of different experimental conditions (perhaps hundreds or thousands of such conditions), what was one data point associated with the expression of one gene under one set of conditions now becomes 106 to 107 data points to be analyzed.

  • Geometric information. Because a great deal of biological function depends on relative shape (e.g., the “docking” behavior of molecules at a potential binding site depends on the three-dimensional configuration of the molecule and the site), molecular structure data are very important. Graphs are one way of representing three-dimensional structure (e.g., of proteins), but ball-and-stick models of protein backbones provide a more intuitive representation.

  • Scalar and vector fields. Scalar and vector field data are relevant to natural phenomena that vary continuously in space and time. In biology, scalar and vector field properties are associated with chemical concentration and electric charge across the volume of a cell, current fluxes across the surface of a cell or through its volume, and chemical fluxes across cell membranes, as well as data regarding charge, hydrophobicity, and other chemical properties that can be specified over the surface or within the volume of a molecule or a complex.

  • Patterns. Within the genome are patterns that characterize biologically interesting entities. For example, the genome contains patterns associated with genes (i.e., sequences of particular genes) and with regulatory sequences (that determine the extent of a particular gene’s expression). Proteins are characterized by particular genomic sequences. Patterns of sequence data can be represented as regular expressions, hidden Markov models (HMMs), stochastic context-free grammars (for RNA sequences), or other types of grammars. Patterns are also interesting in the exploration of protein structure data, microarray data, pathway data, proteomics data, and metabolomics data.

  • Constraints. Consistency within a database is critical if the data are to be trustworthy, and biological databases are no exception. For example, individual chemical reactions in a biological pathway must locally satisfy the conservation of mass for each element involved. Reaction cycles in thermodynamic databases must satisfy global energy conservation constraints. Other examples of nonlocal constraints include the prohibition of cycles in overlap graphs of DNA sequence reads for linear chromosomes or in the directed graphs of conceptual or biological taxonomies.

  • Images. Imagery, both natural and artificial, is an important part of biological research. Electron and optical microscopes are used to probe cellular and organ function. Radiographic images are used to highlight internal structure within organisms. Fluorescence is used to identify the expressions of genes. Cartoons are often used to simplify and represent complex phenomena. Animations and movies are used to depict the operation of biological mechanisms over time and to provide insight and intuitive understanding that far exceeds what is available from textual descriptions or formal mathematical representations.

  • Spatial information. Real biological entities, from cells to ecosystems, are not spatially homogeneous, and a great deal of interesting science can be found in understanding how one spatial region is different from another. Thus, spatial relationships must be captured in machine-readable form, and other biologically significant data must be overlaid on top of these relationships.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement