The paradigm for discovering biological mechanisms is changing.  Classical genetic approaches over the past 50 years have been remarkably successful in finding “the gene for this, and the gene for that”.  However, the utility of these approaches is diminishing as we aim to discover the basis for more complex phenotypes.   A library of strains in which each (non-essential) gene has been knocked out has existed for the model system, S. cerevisiae for quite some time. Much has been learned by experimental queries of this knockout library, yet there is much we still do not know about the cell biology of this organism.  The lesson to be learned is that the mechanisms we seek to understand are not controlled by single genes, but rather by complex networks of interacting genes and gene products.  An additional level of complexity arises as we recognize that biological events arise from dynamic changes in these interaction networks.  Thus, to fully understand biological systems, we must learn the structure and function of the underlying dynamical network mechanisms that control processes at scales from molecular to organismal.

The need to identify network structure/functions brings a new set of challenges to the experimental biologist.  Networks were evolved rather than engineered, and the process of evolution often selects solutions that suffice, rather than solutions that are most efficient.  Like a road system in an old city that “evolved” over time, these networks often appear unnecessarily complex or inefficient.  One of the network features that appears to have selective value is robustness to perturbation.  Biological systems can operate in a wide variety of environmental conditions, so networks appear to have evolved a system to buffer “noise” from changing internal and external environments.  Robustness to perturbation is a nightmare for the experimental geneticist, as genetic perturbations may not produce phenotypic changes.

Our approach has been to develop mathematical and statistical tools for network inference that utilize dynamic measurements.  These inference methods produce network models that prioritize experimental interrogation of the system.  Through an iterative process of modeling and experimentation, the aim is to learn the complex network mechanisms that control biological processes of interest.  We have utilized the highly tractable S. cerevisiae model system to tune our computational and experimental approaches, and we are now deploying these approaches in non-model systems including pathogenic fungi and parasites.