Hypothesis Generation for Genome-scale Metabolic Models
Andrew Stine, Northwestern University
The ease of modern genetic sequencing techniques has produced a plethora of genetic information about a wide variety of organisms. One potential use for this data is the production of genome-scale metabolic models (GSMMs) which are mathematical representations of all the reactions that are known to occur in an organism. GSMMs can be used to computationally elucidate many important characteristics of an organism, such as the minimum growth medium, genetic changes that can increase the production of a compound, and whether a genetic knockout will grow. Unfortunately, all GSMMs contain gaps where chemical compounds are known to be present but the reactions that produce or consume them are not known. This is problematic both because it reduces the accuracy of a GSMM’s predictions and because it highlights shortcomings in our understanding of biology. Previously, researchers have attempted to fill these gaps by hypothesizing missing reactions in a GSMM from a pool of reactions that are known to occur in other organisms. Unfortunately, this technique is not capable of filling all the gaps in GSMMs due to our incomplete knowledge of biological chemistry. Our research group has previously developed a program known as BNICE which utilizes generalized reaction rules to predict probable biological reactions. In this work, we describe a method for the utilization of BNICE to generate hypotheses for likely reactions to fill these gaps. The results of this work will both improve the predictive capabilities of GSMMs and increase our understanding of biological chemistry.
Abstract Author(s): Andrew Stine, Linda Broadbelt