3 March 2017
Michael Savageau
Department of Biomedical Engineering
UC Davis
There is an enormous gap in our ability to relate abundant, rigorously-defined genotype information to the paucity of ad hoc, poorly-defined phenotype information. Bridging this gap, which consists of complex biochemical systems that form the causal mechanistic link connecting genotype and phenotype, is a 'Grand Challenge' and it is widely accepted that predictive mathematical models will play an essential role. Although computationally-based models for large systems provide useful correlations, only experimentally-based models for relatively small sub-systems provide quantitative predictions. Without a fundamental change in modeling strategy, this situation is unlikely to be resolved for the following reasons. There are major difficulties at both experimental and computational levels. Experimentally, there is an enormous number of kinetic parameters for which there is currently little hope of any high-throughput experimental measurement; computationally, the task of exploring the phenotypic repertoire for a given system entails a combinatorial explosion of simulations to sample the high-dimensional space of alternative values for the system parameters, environmental variables, and initial conditions. This is true of all mathematical modeling that employs the conventional 'parameter-centric' strategy involving computer simulation of biochemical systems. We are proposing a fundamentally different 'phenotype-centric' modeling strategy that reverses the conventional 'parameter-centric' paradigm (parameters first – phenotypes last) by first enumerating the repertoire of biochemical phenotypes and then predicting values for the kinetic parameters. The predicted relationships among these parameters offer new possibilities for identifying biological design principles.