Systems biology research starts with a broad question such as ‘How do organelles get built in a cell?’ or ‘What is the gene regulatory circuitry that ensures proper cell differentiation during embryogenesis? To address the question, a model system and set of experimental and computational approaches is chosen. For a model system to be effective, it must be sufficiently complex so that inferences made about its mechanisms and processes will be generalizable, yet sufficiently simple enough that it is amenable to detailed investigation using available technologies. The system must be properly matched to the problem being studied.

Single-cell organisms, such as bacteria and yeast, are popular among systems biologists because of the limited number of parts, in comparison to mammalian systems, and the ease of experimental manipulation. Invertebrates, such as the fruitfly, worm, and sea urchin, are excellent model systems for basic biological processes pertaining to multicellular organisms. Techniques such as wholesale RNAi gene knockouts and comprehensive gene chips are raising perturbation experiments to a new level of sophistication. Some phenomena, such as the presence of an adaptive immune system mediated by MHC and T cells, are present only in vertebrates and must be studied in model organisms such as the mouse or in human cell lines or selected cell populations such as macrophages.

Once the model system and appropriate data collection strategies have been defined, the signature of a systems biology approach is an iterative series of hypothesis-driven exploration and model building endeavors called the “systems biology cycle”. As a “proof of concept,” this process was effectively applied to deciphering the gene-regulatory and protein-interaction networks that underlie pathways of galactose utilization in yeast. Specifically, scientists at the ISB demonstrated the effectiveness of an integrated approach which was used to build, test, and refine a model of a cellular pathway. Perturbations to critical pathway components were analyzed using DNA microarrays, quantitative proteomics, and databases of known physical interactions. Using this approach, 997 messenger RNAs responding to 20 systematic perturbations of the yeast galactose-utilization pathway were identified. Evidence that approximately 15 of 289 detected proteins are regulated posttranscriptionally was obtained, and explicit physical interactions governing the cellular response to each perturbation were delineated. Iterations of network model-building, experimental perturbation, and global measurements prompted and validated hypotheses about the regulation of galactose utilization and the physical interactions that link this to a variety of other metabolic pathways. Several observations were inconsistent with the standard model of galactose utilization generated by more than 30 years of examining genes and proteins one at a time. In several cases, hypotheses were generated to explain these inconsistences, and iterative global perturbations were carried out revealing unexpected additional complexities in the regulation of this system that could not have been delineated by the one gene or protein at a time approach.

As a paradigmatic systems biology project, the yeast galactose utilization analysis relied upon large datasets, integration of different data types, network modeling, and iterative perturbations.

The comprehensive systems biology approach is presented in schematic form. In essence, systems biology aims to progress from descriptive, qualitative models to statistical or probabilistic models that can be used to simulate responses to perturbation of a system’s molecular networks in a way that will yield quantitatively accurate predictions. The development of these modeling tools and the mathematical theories that underlie them constitutes an active area of research for systems biology. (See Pointillist and Dizzy for examples.)

Keywords associated with the methodology of systems biology are:

• Model systems
• Global datasets and analyses
• Integration across multiple data types
• Statistical modeling
• Experimental perturbations
• Iterative hypothesis testing and model building cycles