Knowing how a microbe senses environmental inputs and regulates metabolic changes is important for metabolic engineers trying to direct microbial resources and reactions to specific pathways. Prediction of metabolic changes that result from genetic or environmental perturbations has several important applications, including diagnosing metabolic disorders and discovering novel drug targets. Most of the research in the field of modeling transcriptional regulatory networks (TRNs) and their metabolic effects focuses on integrating metabolic networks with additional data like transcriptional or genomic data. However, these existing methods are limited by the availability of datasets and the huge parameter space associated with TRN models. Thus, there is a need for alternative approaches to modeling regulation of metabolic networks.

It was recently established that microbial cells contain flux sensors which measure the rate at which enzymatic reactions take place, and then adjust, or dial, certain reactions and pathway fluxes. We hypothesize that these flux sensors provide enough information to predict the change in metabolic “dials”, i.e flux splits between different pathways. This project aims to prove the above-mentioned hypothesis using statistical modeling of sensors and dials data in metabolic network simulations.

Using Markov Chain Monte Carlo sampling methods, we sample the flux states of the Escherichia coli K-12 MG1655 strain under varying nutrient sources. We sample from 34 conditions to create a dataset with 340000 datapoints, each representing a unique feasible metabolic flux. We then apply statistical modeling techniques including linear regression, decision trees and ensemble learning methods to predict metabolic dial values using sensor values as input. The results from the statistical modeling techniques show that sensors can effectively predict the dial values without the need for additional data like transcriptional or genomic data.