Periodic Reporting for period 4 - Life-Inspired (Life-inspired complex molecular systems controlled by enzymatic reaction networks)
Okres sprawozdawczy: 2024-04-01 do 2025-03-31
Living organisms have unique capabilities such as self-healing, adaptation to the environment, homeostasis, and converting chemical energy into motion, growth and division. Introducing aspects of autonomous function through sensing of the environment, monitoring the internal state, and regulating behaviour into synthetic life-inspired systems, represents a truly disruptive development, as it challenges our notion of what differentiates living systems from synthetic, man-made devices. However, despite substantial research efforts, a general platform for the construction of such systems remains a highly desired but elusive goal. How do we construct functional systems and devices out of molecules? How do we fuel these systems? How do we replace electronic circuits with networks of chemical reactions?
The ultimate aim of this proposal is to construct life-inspired complex molecular systems based on the design blueprints of living matter. Achieving this aim would yield life-like materials with embedded computing power that have the ability to sense their environment, to compute information from the environment, and to learn and adapt their shape and function. Such materials might become a radically new interface between electronic and living systems.
This project has been very successful - in an unexpected way. Instead of a systematic, bottom-up approach to networks of increasing complexity, we discovered that self-organizing chemical reaction networks have so-called reservoir computation capabilities. We demonstrated that both the formose reaction as well as a novel type of enzymatic networks based on molecular competition can outperform in silico machine learning algorithms on complex tasks such as non-linear classification and time-series prediction. Furthermore, as computation is done in the chemical domain we could demonstrate that such systems can act as reservoir sensors, that classify the chemical characteristics of their environment. We fully achieved our goal of establishing a blueprint for smart materials that can interact with both electronic and living systems.
The ultimate aim of this proposal is to construct life-inspired complex molecular systems based on the design blueprints of living matter. Achieving this aim would yield life-like materials with embedded computing power that have the ability to sense their environment, to compute information from the environment, and to learn and adapt their shape and function. Such materials might become a radically new interface between electronic and living systems.
This project has been very successful - in an unexpected way. Instead of a systematic, bottom-up approach to networks of increasing complexity, we discovered that self-organizing chemical reaction networks have so-called reservoir computation capabilities. We demonstrated that both the formose reaction as well as a novel type of enzymatic networks based on molecular competition can outperform in silico machine learning algorithms on complex tasks such as non-linear classification and time-series prediction. Furthermore, as computation is done in the chemical domain we could demonstrate that such systems can act as reservoir sensors, that classify the chemical characteristics of their environment. We fully achieved our goal of establishing a blueprint for smart materials that can interact with both electronic and living systems.
The project has yielded a number of important technological and scientific advances. In our efforts to construct synthetic enzymatic reaction networks, we established a robust method for immobilizing enzymes on microfluidically prepared beads. This allowed us to construct networks by placing beads in a flow reactor. Over the course of this proposal, we expanded the compleity of the network from just 2-3 enzymes to 12 enzymes comprising most of the glycolysis pathway. As these networks showed increasingly complex dynamics, we developed mass spectrometry methods to observe all intermediates in a quantitative way and used Bayesian inference to rapidly extract kinetic information. Our platform for extracting kinetic information has become ever more sophisticated and we successfully applied for an ERC PoC grant to explore potential commercialisation. We ultimately concluded that valorisation was too early, but we will use the platform in future grant applications to work towards the optimization of networks that produce valuable compounds.
In a separate strand, we studied the potential of reaction networks for novel types of computation. We discovered that both the formose reaction as well as a novel type of enzymatic networks based on molecular competition have so-called reservoir computation capabilities can outperform in silico machine learning algorithms on complex tasks such as non-linear classification and time-series prediction. As enzymes are sensitive to pH and temperature, we could demonstrate that the reservoirs also had sensing capabilities. We fully achieved our goal of establishing a blueprint for smart materials that can interact with both electronic and living systems.
The work has resulted in 2 PhD thesis that have already been defended (with three more pending), and 9 publications in peer reviewed journals ( including in leading journals such as Nature, Nature Communications (2x), Journal of the American Chemical Society and Angewandte Chemie).
In a separate strand, we studied the potential of reaction networks for novel types of computation. We discovered that both the formose reaction as well as a novel type of enzymatic networks based on molecular competition have so-called reservoir computation capabilities can outperform in silico machine learning algorithms on complex tasks such as non-linear classification and time-series prediction. As enzymes are sensitive to pH and temperature, we could demonstrate that the reservoirs also had sensing capabilities. We fully achieved our goal of establishing a blueprint for smart materials that can interact with both electronic and living systems.
The work has resulted in 2 PhD thesis that have already been defended (with three more pending), and 9 publications in peer reviewed journals ( including in leading journals such as Nature, Nature Communications (2x), Journal of the American Chemical Society and Angewandte Chemie).
Our work on analyzing complex network dynamics has progressed far beyond the state of the art, and the postdoc involved in the project is now taking this topic further in his new position at the University of Amsterdam. By incorporating machine learning/AI algorithms, he is now looking at even more complex networks and develop methodologies that can learn the underlying network structure by observing the dynamic response of the network to external perturbations.
Our work on in chemico reservoir computing is completely novel and has laid the foundation for a new approach to molecular computing. We will certainly pursue this topic in future research projects.
Our work on in chemico reservoir computing is completely novel and has laid the foundation for a new approach to molecular computing. We will certainly pursue this topic in future research projects.