The Possibility of Automated Experiments for Inference of Metabolic Models

By John Wikswo, Vanderbilt University

Sep 16, 2008

CNLS Conference room.

Recent advances in systems biology and metabolomics have been enabled by high throughput analyses of gene expression and metabolites, clever experiments, and computer modeling that allow step-by-step dissection of cellular metabolic and signaling pathways. However, many of the techniques perfected for such analyses are not readily applicable to determination of the fast components of cellular networks. The de novo inference of a metabolic network is daunting task in reverse engineering of a complex and highly non-linear dynamic system, particularly if attention is paid to the transient response. The automated analysis of fast cellular control and signaling mechanisms requires 1) micro- and nanoinstruments that measure simultaneously multiple extracellular and intracellular variables with sufficient bandwidth to resolve the sub-second dynamics of the transient departure from homeostasis; 2) the ability to open and close at will existing internal control and signaling loops; 3) microactuators such as BioMEMS valves, pumps, and electrodes that provide high bandwidth feedback to the external environment, 4) externally addressable nanoactuators that operate intracellularly; and 5) inverse metabolic and signaling models and control algorithms to characterize the physiological and pathological effects of a variety of interventions. Such an approach would enable the rapid determination of the immediate response of cells to chemical and biological agents and environmental toxins, which is critical for both the timely classification of unknown or hacked agents and the identification of appropriate therapies and prophylaxes. We demonstrate that simultaneous measurement of the dynamic metabolic response of small populations of cells can discriminate within minutes between various Class A CBW toxins and agents. The challenge is to devise the instruments, models, and protocols required to extend this approach to dynamic analysis of multiple variables in selected metabolic or signaling pathways. We believe that self-learning bioreactors programmed to perform symbolic regression and optimized experimental design may provide a valuable key to unlocking the complexities of metabolic network dynamics, detecting the presence or effects of CBW agents, discovering new drugs, and identifying undesirable side effects of existing ones.

Back to CNLS q-bio Seminars.