Search

Cellular metabolism is a highly interconnected network with thousands of reactions that convert nutrients into the molecular building blocks of life. Metabolic connectivity varies greatly with cellular context and environmental conditions, and it remains a challenge to compare genome-scale metabolism across cell types because of the high dimensionality of the reaction flux space. Here, we employ self-supervised learning and genome-scale metabolic models to compress the flux space into low-dimensional representations that preserve structure across cell types. We trained variational autoencoders (VAEs) on large fluxomic data (N = 800, 000) sampled from patient-derived models for various cancer cell types. The VAE embeddings have an improved ability to distinguish cell types than the uncompressed fluxomic data, and sufficient predictive power to classify cell types with high accuracy. We tested the ability of these classifiers to assign cell type identities to unlabelled patient-derived metabolic models not employed during VAE training. We further employed the pre-trained VAE to embed another 38 cell types and trained multilabel classifiers that display promising generalization performance. Our approach distils the metabolic space into a semantically rich vector that can be used as a foundation for predictive modelling, clustering or comparing metabolic capabilities across organisms.

The identification of essential genes, i.e. those that impair cell survival when deleted, requires large growth assays of knock-out strains. The complexity and cost of such experiments has triggered a growing interest in computational methods for prediction of gene essentiality. In the case of metabolic genes, Flux Balance Analysis (FBA) is widely employed to predict essentiality under the assumption that cells maximize their growth rate. However, this approach assumes that knock-out strains optimize the same objectives as the wild-type, which excludes cases in which deletions cause large physiological changes to meet other objectives for survival. Here, we resolve this limitation with a novel machine learning approach that predicts essentiality directly from wild-type flux distributions. We first project the wild-type FBA solution onto a mass flow graph, a digraph with reactions as nodes and edge weights proportional to the mass transfer between reactions, and then train binary classifiers on the connectivity of graph nodes. We demonstrate the efficacy of this approach using the most complete metabolic model of Escherichia coli, achieving near state-of-the art prediction accuracy for essential genes. Our approach suggests that wild-type FBA solutions contain enough information to predict essentiality, without the need to assume optimality of deletion strains.

Metabolic engineering has enabled the production of a wealth of chemicals with microorganisms. Classic strategies for pathway engineering rely on the expression of heterologous enzymes in a host that convert native intermediates into target products. Although traditional implementations are based on open-loop control, recent advances in gene circuit engineering offer opportunities for building feedback systems that dynamically control pathway activity. Here we present a framework for the design of metabolic control circuits based on multi objective optimization. We show that positive and negative feedback loops produce a range of optimal dynamics along a Pareto front. Such regulatory loops define connectivities between pathway intermediates and enzymatic genes that trade-off metabolic production against the burden to the host. Our results lay the groundwork for the automated design of gene circuitry in applications at the interface of synthetic biology and metabolic engineering.

Mathematical models are powerful tools for the design of gene circuits in synthetic biology. Models can predict circuit behavior and guide the selection of suitable circuit components and implementations. The increasing complexity synthetic circuits typically requires model approximations that are amenable for analysis and parameter estimation. However, these simpler models may not reflect the actual behavior of a synthetic circuit once implemented in a host of interest. The experimental implementation of a circuit may show dynamics that cannot be accounted by simplified models. Potential sources for unexpected circuit behavior are host-circuit interactions through competition for shared cellular resources. The allocation of cellular resources to circuit expression inevitably reduces those resources available to other cellular functions such as growth and biosynthesis. Yet the balance between circuit function and host physiology is often overlooked at the design stage. Here we integrate a model of a synthetic circuit, the incoherent feedforward loop, with a mechanistic model for cell growth. We show how the integrated model explains measured circuit responses that the circuit-only model was unable to predict. Our results provide evidence that anomalous circuit dynamics can be traced back to host-circuit interactions.

Genetic feedback is one of the mechanisms that enables metabolic adaptations to environmental changes. The stable equilibria of these feedback circuits determine the observable metabolic phenotypes. We consider an unbranched metabolic network with one metabolite acting as a global regulator of enzyme expression. Under switch-like regulation and exploiting the time scale separation between metabolic and genetic dynamics, we develop geometric criteria to characterize the equilibria of a given network. These results can be used to detect mono- and bistability in terms of the gene regulation parameters for any combination of activation and repression loops.