Chemical programming to exploit chemical reaction systems for computation

All known life forms process information on molecular level. Biochemical information processing found in nature is known to be robust, self-organizing, adaptive, decentralized, asynchronous, fault-tolerant, and evolvable. In order to further exploit these properties new programming techniques are required. This thesis is on programming approaches to exploit the computational capabilities of chemical systems, consisting of two parts. In the first part, constructive design, research activities on theoretical development of chemical programming are reported. As results of the investigations, general programming principles, named organizationoriented programming, are derived. The idea is to design reaction networks such that the desired computational outputs correspond to the organizational structures within the networks. Due to a relation between organizations and fixed points, supported by a theorem, the constructed reaction networks can be utilized to develop dynamical reaction systems with desired output behaviors. The second part, autonomous design, discusses on programming strategies without human interactions, namely evolution and exploration. Motivations for this programming approach include possibilities to discover novelty without rationalization. Regarding first the evolutionary strategies, we rather focused on how to track the evolutionary processes. Our approach is to analyze these dynamical processes on a higher level of abstraction, and usefulness of distinguishing organizational evolution in space of organizations from actual evolution in state space is emphasized. As second strategy of autonomous chemical programming, we suggest an explorative approach, in which an automated system is utilized to explore the behavior of the chemical reaction system as a preliminary step. A specific aspect of the system’s behavior becomes ready for a programmer to be chosen for a particular computational purpose. In this thesis, developments of autonomous exploration techniques are reported. Finally, we discuss combining those two approaches, constructive design and autonomous design, titled as a hybrid approach. From our perspective, hybrid approaches are ideal, and cooperation of constructive design and autonomous design is fruitful. In order to design chemical computing systems, those approaches can be employed alternatively from developing modular parts of chemical programs to developing whole programs. Furthermore, constructing chemical reaction systems produce useful experiences to improve autonomous designs. Autonomously designed systems are proper sources of inspiration for chemical programming in a constructive way.