Periodic Reporting for period 1 - BiosenSAI (Biosensing by Sequence-based Activity Inference)
Période du rapport: 2024-02-01 au 2026-07-31
The ability of cells to sense and respond to signals is an essential requirement of life. Genetically encoded biosensors meet this need by detecting, for example, chemicals and triggering gene expression in response. This concept is used across the life sciences to sense molecules in basic research, diagnostics, and treatment. Crucially, biosensors can be used to isolate and engineer microbes that sustainably produce value-added chemicals from renewable starting materials and thus play a key role in the transition to a circular economy. For instance, they can be used to find the “best producers” in large pools of natural or genetically engineered microbes, which is in many cases the bottleneck in the development of new biotechnological processes and products.
However, native biosensors are usually unfit for most of the desired applications, since they do not sense the right molecules (or products) of interest and they frequently do not respond to the right concentration range. This project aims at overcoming this limitation using a data-driven engineering approach. It involves the development of novel methods to experimentally assess biosensor variants in extremely high numbers (up to hundreds of millions per experiment) at low experimental cost and effort. Furthermore, it entails the exploitation of the resulting “big data” on biosensors with cutting-edge machine learning techniques to build computer models for the design of biosensors. The overall goal of the project is the development of an integrated platform for the engineering and design of biosensors with new-to-nature properties “à la carte”. This novel, data-driven approach aims at breaking new grounds in biosensor engineering through synergies between synthetic biology and artificial intelligence paving the way to novel, sustainable bioprocesses.
However, native biosensors are usually unfit for most of the desired applications, since they do not sense the right molecules (or products) of interest and they frequently do not respond to the right concentration range. This project aims at overcoming this limitation using a data-driven engineering approach. It involves the development of novel methods to experimentally assess biosensor variants in extremely high numbers (up to hundreds of millions per experiment) at low experimental cost and effort. Furthermore, it entails the exploitation of the resulting “big data” on biosensors with cutting-edge machine learning techniques to build computer models for the design of biosensors. The overall goal of the project is the development of an integrated platform for the engineering and design of biosensors with new-to-nature properties “à la carte”. This novel, data-driven approach aims at breaking new grounds in biosensor engineering through synergies between synthetic biology and artificial intelligence paving the way to novel, sustainable bioprocesses.
In the first reporting period of this project, we made critical development steps for the required experimental technology. To this end, we have successfully developed the required molecular biology tools (e.g. plasmids) and experimental protocols needed to characterize extremely large numbers of variants for RNA- and protein-based biosensors. The successful outcome was demonstrated in a set of experiments, in which genetic variant libraries of model biosensors were characterized by measuring their response towards a wide range of biotechnologically relevant chemicals including bulk chemicals, flavour and fragrance molecules and pharmaceutically active compounds. Furthermore, we have developed computational tools to process and analyse the resulting big datasets, which in some cases exceeded hundreds of millions individual data points. Despite the still early stage of the project, it has already delivered a set of critical proofs of concept confirming the validity of the approach and the high potential for a wide range of applications across the life sciences.
Note that due to a change in the host institution of the PI, funding through the ERC was discontinued after 17 months and the project was continued at the new host institution.
Note that due to a change in the host institution of the PI, funding through the ERC was discontinued after 17 months and the project was continued at the new host institution.
Despite the early stage of the project, it has already delivered significant progress beyond the state of the art in biosensor engineering and design. The newly developed methodology allows access to experimental datasets on biosensors of previously inaccessible size (hundreds of millions of data points) and quality (e.g. high dynamic resolution, low experimental noise). Furthermore, it has delivered a set of candidate biosensors with new-to-nature properties, which are for instance able to sense new target molecules and are currently further investigated. In particular, the access to such “big data” on biosensors has the potential to shed light into a critical question: How to changes in the genetic sequence of a biosensor affect its sensory properties? This will not only contribute to a significantly improved fundamental understanding of how biosensors work but allow to rationalize and streamline their engineering through machine learning-guided approaches that minimize development effort and maximize the chances of successful outcomes.
Note that due to a change in the host institution of the PI, funding through the ERC was discontinued after 17 months and the project was continued at the new host institution.
Note that due to a change in the host institution of the PI, funding through the ERC was discontinued after 17 months and the project was continued at the new host institution.