Mathilde Koch

Modélisation pour la conception et l’analyse de circuits métaboliques synthétiques: Les buts de cette thèse sont doubles, et concernent les circuits métaboliques synthétiques, qui permettent de détecter des composants chimiques par transmission de signal et de faire du calcul en utilisant des enzymes. La première partie a consisté à développer des outils d’apprentissage actif et par renforcement pour améliorer la conception de circuits métaboliques et optimiser la biodétection et la bioproduction. Pour atteindre cet objectif, un nouvel algorithme (RetroPath3.0) fondé sur une recherche arborescente de Monte Carlo guidée par similarité est présenté. Cet algorithme, combiné à des règles de réaction apprises sur des données et des niveaux différents de promiscuité enzymatique, permet de focaliser l’exploration sur les composés et les chemins les plus prometteurs en bio-rétrosynthèse. Les chemins obtenus par rétrosynthèse peuvent être implémentés dans des cellules ou des systèmes acellulaires. Afin de concevoir le meilleur milieu pour optimiser la productivité du système, une méthode d’apprentissage actif qui explore efficacement l’espace combinatoire des composants du milieu a été développée.La deuxième partie a consisté à développer des méthodes d’analyse, pour générer des connaissances à partir de données biologiques, et modéliser les réponses de biocapteurs. Dans un premier temps, l’effet du nombre de copies de plasmides sur la sensibilité d’un biocapteur utilisant un facteur de transcription a été modélisé. Ensuite, en utilisant des systèmes acellulaires qui permettent un meilleur contrôle des variables expérimentales comme la concentration d’ADN, l’utilisation des ressources a été modélisée pour assurer que notre compréhension actuelle des phénomènes sous-jacents est suffisante pour rendre compte du comportement du circuit, en utilisant des modèles empiriques ou mécanistiques. Couplés aux outils de conception de circuits métaboliques, ces modèles ont ensuite permis de développer une nouvelle approche de calcul biologique, appelée perceptrons métaboliques.Dans l’ensemble, cette thèse présente des outils de conception et d’analyse pour les circuits métaboliques synthétiques. Ces outils ont été utilisés pour développer une nouvelle méthode permettant d’effectuer des calculs en biologie synthétique.

Metabolic engineering aims to produce chemicals of interest from living organisms, to advance towards greener chemistry. Despite efforts, the research and development process is still long and costly and efficient computational design tools are required to explore the chemical biosynthetic space. Here, we propose to explore the bio-retrosynthesis space using an Artificial Intelligence based approach relying on the Monte Carlo Tree Search reinforcement learning method, guided by chemical similarity. We implement this method in RetroPath RL, an open-source and modular command line tool. We validate it on a golden dataset of 20 manually curated experimental pathways as well as on a larger dataset of 152 successful metabolic engineering projects. Moreover, we provide a novel feature, that suggests potential media supplements to complement the enzymatic synthesis plan.

Lysate-based cell-free systems have become a major platform to study gene expression but batch-to-batch variation makes protein production difficult to predict. Here we describe an active learning approach to explore a combinatorial space of ~4,000,000 cell-free compositions, maximizing protein production and identifying critical parameters involved in cell-free productivity. We also provide a one-step-method to achieve high quality predictions for protein production using minimal experimental effort regardless of the lysate quality.

Synthetic biological circuits are promising tools for developing sophisticated systems for medical, industrial, and environmental applications. So far, circuit implementations commonly rely on gene expression regulation for information processing using digital logic. Here, we present a different approach for biological computation through metabolic circuits designed by computer-aided tools, implemented in both whole-cell and cell-free systems. We first combine metabolic transducers to build an analog adder, a device that sums up the concentrations of multiple input metabolites. Next, we build a weighted adder where the contributions of the different metabolites to the sum can be adjusted. Using a computational model fitted on experimental data, we finally implement two four-input perceptrons for desired binary classification of metabolite combinations by applying model-predicted weights to the metabolic perceptron. The perceptron-mediated neural computing introduced here lays the groundwork for more advanced metabolic circuits for rapid and scalable multiplex sensing.

Cell-free transcription–translation systems have great potential for biosensing, yet the range of detectable chemicals is limited. Here we provide a workflow to expand the range of molecules detectable by cell-free biosensors through combining synthetic metabolic cascades with transcription factor-based networks. These hybrid cell-free biosensors have a fast response time, strong signal response, and a high dynamic range. In addition, they are capable of functioning in a variety of complex media, including commercial beverages and human urine, in which they can be used to detect clinically relevant concentrations of small molecules. This work provides a foundation to engineer modular cell-free biosensors tailored for many applications.

Cell-free TX-TL is an increasingly mature and useful platform for prototyping, testing, and engineering biological parts and systems. However, to fully accomplish the promises of synthetic biology, mathematical models are required to facilitate the design and predict the behavior of biological components in cell-free extracts. We review here the latest models accounting for transcription, translation, competition, and depletion of resources as well as genome scale models for lysate-based cell-free TX-TL systems, including their current limitations. These models will have to find ways to account for batch-to-batch variability before being quantitatively predictive in cell-free lysate-based platforms.

Cell-free transcription-translation systems have great potential for biosensing, yet the range of detectable chemicals is limited. Here we provide a framework to expand the range of molecules detectable by cell-free biosensors by combining synthetic metabolic cascades with transcription factor-based networks. These hybrid cell-free biosensors are highly-sensitive and have a fast response and high-dynamic range. This work provides a foundation to engineer modular cell-free biosensors tailored for many applications.

Progress in synthetic biology tools has transformed the way we engineer living cells. Applications of circuit design have reached a new level, offering solutions for metabolic engineering challenges that include developing screening approaches for libraries of pathway variants. The use of transcription‐factor‐based biosensors for screening has shown promising results, but the quantitative relationship between the sensors and the sensed molecules still needs more rational understanding. Herein, we have successfully developed a novel biosensor to detect pinocembrin based on a transcriptional regulator. The FdeR transcription factor (TF), known to respond to naringenin, was combined with a fluorescent reporter protein. By varying the copy number of its plasmid and the concentration of the biosensor TF through a combinatorial library, different responses have been recorded and modeled. The fitted model provides a tool to understand the impact of these parameters on the biosensor behavior in terms of dose–response and time curves and offers guidelines to build constructs oriented to increased sensitivity and or ability of linear detection at higher titers. Our model, the first to explicitly take into account the impact of plasmid copy number on biosensor sensitivity using Hill‐based formalism, is able to explain uncharacterized systems without extensive knowledge of the properties of the TF. Moreover, it can be used to model the response of the biosensor to different compounds (here naringenin and pinocembrin) with minimal parameter refitting.