The evolution of metabolism regulation is an intertwined process, where different strategies are constantly being developed towards a cognitive ability to perceive and respond to an environment. Organisms depend on an orchestration of a complex set of chemical reactions: maintaining homeostasis with a changing environment, while simultaneously sending material and energetic resources to where they are needed. The success of an organism requires efficient metabolic regulation, highlighting the connection between evolution, population dynamics and the underlying biochemistry. In this work, I represent organisms as coupled information-processing networks, that is, gene-regulatory networks receiving signals from the environment and acting on chemical reactions, eventually affecting material flows. I discuss the mechanisms through which metabolism control is improved during evolution and how the nonlinearities of competition influence this solution-searching process. The propagation of the populations through the resulting landscapes generally point to the role of the rhythm of cell division as an essential phenotypic feature driving evolution. Subsequently, as it naturally follows, different representations of organisms as oscillators are constructed to indicate more precisely how the interplay between competition, maturation timing and cell-division synchronisation affects the expected evolutionary outcomes, not always leading to the \"survival of the fastest\".

Für die erfolgreiche Lösung eines Optimierungsproblems ist die Wahl der verwendeten Suchstrategie von entscheidender Bedeutung. Die vorliegende Arbeit untersucht die Kriterien dieser Wahl. Dabei stellen sich drei grundlegende Fragen: Welche Strategien der Optimierung eines gegebenen Problems existieren überhaupt, und was für Eigenschaften besitzen sie? Wodurch wird der Charakter eines Optimierungsproblems bestimmt, und gibt es Klassen ähnlicher Probleme? Besteht eine Verbindung zwischen den Eigenschaften der Strategien und den Klassen der Probleme, die es ermöglicht, für jede Problemklasse eine geeignete Optimierungsstrategie anzugeben? Dazu wird zuerst die Klasse der Evolutionären Algorithmen naher betrachtet, deren generelles Verhalten die Boltzmannstrategie, Darwinstrategie oder Boltzmann-Darwin-Strategie beschreiben. Als weiteres Beispiel wird die Multitome Strategie untersucht. In ihr wird das Problem unter verschiedenen Gesichtspunkten betrachtet und in Einzelanforderungen zerlegt, die abwechselnd optimiert werden. Für den speziellen Fall der Dichotomen Strategie wird die allgemeine zeitabhängige Lösung mit Hilfe der Methode der Charakteristiken bestimmt. Zur Beantwortung der zweiten Frage wird die Zustandsdichte als klassifizierende Größe des Optimierungsproblems eingeführt. Sie kann unter Verwendung der Boltzmannstrategie während des Optimierungslaufes durch zwei allgemeine Approximationsmethoden: die Methode der stationären Verteilungen und die Eigenvektormethode bestimmt werden. Aus der Zustandsdichte erhält man den Wirkungsgrad der Zufallssuche. Er charakterisiert den Ordnungsgrad des Problems und stellt damit ein wichtiges Maß der Problemschwierigkeit dar. Die entscheidende dritte Frage wird für Probleme der Optimierung frustrierter Sequenzen, der Netzwerkoptimierung und für das Faltungsproblem der RNA behandelt. Mit der Einführung der Klassen gerichteter und ungerichteter Strategien, die für Optimierungsprobleme mit niedrigem bzw. hohem Wirkungsgrad der Zufallssuche effektiv sind, kann eine Verbindung zwischen dem Strategieverhalten und dem Problemcharakter hergestellt werden, die es ermöglicht, für eine konkrete Optimierungsaufgabe die Klasse der geeigneten Strategien zu wählen. / A crucial point of successful solving an optimization problem is the choice of the used strategy. The present paper investigates the criteria of this choice. Thereby three fundamental questions put themselves: Which strategies of the optimization of a given problem exist altogether, and which properties characterize the strategies? How is the character of an optimization problem determined, and are there classes of similar problems? Does a combination exist between the characteristics of the strategies and the classes of problems, which makes it possible to indicate a suitable strategy for each class? The class of the Evolutionary Algorithms is considered more closely. The general behavior of the algorithms can be described by the Boltzmann strategy, Darwin strategy or Boltzmann-Darwin strategy. As a further example the Multitomic strategy is explored. In this approach the problem is considered under different points of view and decomposes in single demands, which are optimized alternately. For the special case of the Dichotomic strategy the general time dependent solution is determined. To answer the second question the density of states is introduced as classifying measure of optimization problems. The density can be determined during the optimization course by two general approaches: the method of the stationary distribution and the eigenvalue method. From the density of states one receives the efficiency of the random search. It describes the degree of order of the problem and presents an measure of the problem difficulty. The important third question is treated for problems of the optimization of frustrated sequences, the network optimization and RNA folding. The introduction of the classes of directed and non directed strategies, which are effective for problems with low and high efficiency of the random search, establishes a connection between the strategy and the character of the problem, which makes it possible to choose the class of the suitable strategies for a given optimization task.

The evolution of metabolism regulation is an intertwined process, where different strategies are constantly being developed towards a cognitive ability to perceive and respond to an environment. Organisms depend on an orchestration of a complex set of chemical reactions: maintaining homeostasis with a changing environment, while simultaneously sending material and energetic resources to where they are needed. The success of an organism requires efficient metabolic regulation, highlighting the connection between evolution, population dynamics and the underlying biochemistry. In this work, I represent organisms as coupled information-processing networks, that is, gene-regulatory networks receiving signals from the environment and acting on chemical reactions, eventually affecting material flows. I discuss the mechanisms through which metabolism control is improved during evolution and how the nonlinearities of competition influence this solution-searching process. The propagation of the populations through the resulting landscapes generally point to the role of the rhythm of cell division as an essential phenotypic feature driving evolution. Subsequently, as it naturally follows, different representations of organisms as oscillators are constructed to indicate more precisely how the interplay between competition, maturation timing and cell-division synchronisation affects the expected evolutionary outcomes, not always leading to the \"survival of the fastest\".