Fig. 2
From: A machine learning approach to predict metabolic pathway dynamics from time-series multiomics data
Cycle for learning metabolic system dynamics from time-series proteomics and metabolomics data. (1) Experimentally, time-series proteomics and metabolomics data are acquired for several strains of interest (represented by different colors). These data are represented in a metabolomics phase space, with an axis corresponding to each measured metabolite. (2) The time-series data traces are smoothed and differentiated (Supplementary Fig. S2). The derivatives provide the training data to derive the relationship between metabolomics and proteomics data and the metabolite change (Supplementary Fig. S1, Eq. (1)). (3) The state derivative pairs are fed into a supervised machine learning algorithm. The machine learning algorithm learns and generalizes the system dynamics from the examples provided by each strain. (4) The model can then be used to simulate virtual strains and explore the metabolic space looking for mechanistic insight or commercially valuable designs. This process can then be repeated using the model to create new strains, which will further improve the accuracy of the dynamic model in the next round
