Analysing network motifs is a common way of characterising biological networks. Motifs are small subgraphs that are more abundant in the observed network than would be expected in random graphs. They may play an important role in network function, and as such may be selected by evolution. In some cases, such as neural networks, they are instantiated via a developmental process. The processes used to structure Artificial Neural Networks, whether training or evolution, do not usually result in motifs or modularity more generally. We introduce a new version of Developmental Graph Cellular Automata (DGCA) which can be used in an evolutionary and developmental (evo-devo) process to produce networks with specific motif profiles. We evolve developmental rules (the “genome”) so that networks are produced with similar motif profiles to specific biological networks. Networks produced in this way may have useful computational and/or dynamical properties when deployed as Recurrent Neural Networks (RNNs) or in Reservoir Computing (RC).

When building models that simulate biological systems at different levels of abstraction, we need to compare their output parameters with measurements from lab experiments. Before we can do that, we need a way to parameterise the models themselves. Here, we investigate parameterising an abstract computational model of plasmid circuits operated by DNA supercoiling (TORCComp), using a more detailed biophysical model (TORCPhys). TORCComp is built as a high speed low fidelity model, which will allow us to explore many variations of parameters in our modelled systems, at the higher abstract level of circuit components. This is aimed at increasing our ability to design supercoiling operated plasmid circuits. TORCPhys is a slower more detailed model, whose parameters are derived from physical concepts and lab experiments, designed to simulate the detailed action of a single circuit at the lower biomolecular level. It cannot be used as an exploratory tool for circuit construction due to longer run times. To explore the feasibility of using TORCPhys to parameterise TORCComp, here, we compare the models of a simple supercoiling controlled plasmid circuit operational in bacteria ( Escherichia coli or Salmonella enterica ) through the mappings of their states. We parameterise TORCComp based on parameter values that are physiologically observable in both lab experiments and TORCPhys, and also those that are not observable in the lab, but can be observed in TORC-Phys. Our results demonstrate the difficulty of parameterising a model based on limited highly contextual observations, and the difficulty of comparing models at different abstraction levels.

We explore a wide variety of behaviours possible with Developmental Graph Cellular Automata. We use novelty search to find more extreme types of behaviour in terms of transient length and attractor cycle length. This also serves as a proof-of-concept that the system is evolvable. We then examine in more detail some individual examples of interesting behaviour, particularly focusing on cases where the graph divides into two or more separate components.

Synthetic biology is one facet of Artificial Life which designs novel biological components, e.g. DNA, RNA, membranes, to produce new behaviours. Here, we are interested in DNA “circuits”: DNA engineered to have particular computational properties. During gene transcription, the DNA double-helix undergoes supercoiling changes, which affects transcription of nearby genes. There is limited mathematical, as opposed to physical, modelling of DNA circuits, and supercoiling is not considered. In many current synthetic circuits, supercoiling has to be carefully removed, particularly in in vivo systems, to prevent unmodelled side effects. However, supercoiling is an intrinsic property of DNA that impacts gene expression, and could be exploited if included in models. Here, we present a new π -calculus formalism for modelling DNA circuits with supercoiling, and demonstrate its use on a simple genetic circuit. The state transition diagrams normally associated with π -calculus are not accessible when the number of states becomes large. We present a new circular visualisation of the π -calculus circuit components that is more intuitive and readable for biologists familiar with the circular visualisations of plasmids.

We present a system for growing graphs which can be thought of as an extension of the update rules used by Cellular Automata. As in Neural Cellular Automata, these rules are encoded in the real-valued weight matrix of a neural network. This should make the system easy to evolve, allowing it to be used as an evolutionary-developmental method of creating graph structures for use as recurrent neural networks or substrates in Reservoir Computing. Here we conduct a random search experiment and characterise five different classes of behaviour of the system. The most interesting of these is when the graph grows for a number of timesteps before naturally coming to a halt as it enters an attractor. This behaviour is seen more frequently than might be expected and contrasts with most developmental systems in which growth must be stopped by external intervention. There are clear parallels with biological morphogenetic processes where growth naturally comes to a halt.

Open-ended novelty is one of the goals of ALife. This provides challenges for analysis as the system evolves. We provide definitions for several emergent properties, such as parasitism and hypercycles, observed to emerge in an RNA world configuration of the Stringmol automata chemistry, and show how these can simultaneously be mathematically simple, capture the complexity of the processes, and be readily implementable.

Evolutionary algorithms are powerful tools to discover novel and diverse solutions to complex problems. Here, we discuss how open-ended algorithms, such as novelty search, can be used to design and evaluate new unconventional computing systems, from the design of materials to the creation of new computational models.

We present a new modelling approach for complex systems incorporating a dynamic environment and individuals with agency. We do this through multiple models at different levels. We develop a common meta-model for these kinds of models. The meta-model captures the concepts of agents moving and interacting on a dynamic network, to provide the power of an agent based model situated in the context of a dynamic and changing environment. The addition of context allows us to isolate the decision process of the agent from the constraints and resources provided by the environment, so we can consider separately the effect of changes in the environment from changes in the agents’ decision process, and changes caused by agents acting differently based on their learning from, and adapting to, the changed environment. We develop a generalised platform model for implementing different complex systems conforming to the meta-model. We illustrate the approach by developing a domain model for a particular system of interest, a simplified model of declining mobility, which we use to guide the specialisation of the generic platform model to an implementation and to perform simulation experiments.

Open-ended novelty is one of the goals of ALife. We use a recent definition of open-endedness, stated in terms of system models and meta-models, to demonstrate how the Stringmol Automata Chemistry achieves variation, innovation and emergence in a replicator-parasite system. We also show how Stringmol's self-modifying code allows certain of these novelties to be exploited within the system itself, while others are only externally observed.

We introduce a new sub-symbolic Artificial Chemistry, called the Meta-Atom Artificial Chemistry. It treats composite particles (composites of random boolean networks, RBN) as a new type of higher level atom, a meta-atom . These complex structures, together with a new kind of link, then form even larger, multi-level, structures. We show that Meta-Atom Artificial Chemistry exhibits rich behaviour, including reaction pathways that resemble catalytic reactions.

Our MetaChem framework supports the definition and combination of artificial chemistries. Here we describe an implementation of MetaChem in an object oriented language. We briefly define MetaChem, and provide an example in the form of a toy AChem: StringCatChem. We present the class hierarchy used to define MetaChem such that the implementation can run directly from a graph description of some AChem. This matches the description given by the formal framework definition. We also describe some generic functions of MetaChem that have been implemented and used in StringCatChem. This implementation is available on GitHub.

We introduce a modularisation of artificial chemistries (AChems). This allows us to define a standard linking method between AChems. We illustrate the approach with a system that nests a Jordan Algebra AChem (JA AChem) inside agents of SwarmChem, and show how our modular approach allows us to define and experiment with multiple variants in a standard manner. Potential for future formalisation is discussed.

Natural chemistry deals with non-deterministic processes, and this is reflected in some artificial chemistries. We can tune these artificial systems by manipulating the functions that define their probabilistic processes. In this work we consider different probabilistic functions for particle linking, applied to our Jordan Algebra Artificial Chemistry. We use five base functions and their variations to investigate the possible behaviours of the system, and try to connect those behaviours to different traits of the functions. We find that, while some correlations can be seen, there are unexpected behaviours that we cannot account for in our current analysis. While we can set and manipulate the probabilities in our system, it is still complex and still displays emergent behaviour that we can not fully control.

We explore the effects that different reactor types have on Spiky-RBN AChem systems, looking at mass conserving and flow reactors. To assist in analysing the behaviour we introduce an activity measure based on possible system state changes as a result of changes in particle properties. This leads to a discussion on approaches to engineering complex systems towards specific goals.

Computational reflection uses software architectures that are capable of self- modification at runtime. These systems have implementations between two extremes: procedural reflection, in which unlimited self-modification is available at the expense of infinite recursion; and declarative reflection, which uses pre-defined metrics to drive the self-modification and is hence limited in scope. Biological processes also exploit the concept of reflection, where natural selection drives the process of modification. The concept of a program in computing has an analogy with an individual member of a species. The process of life is discretised into a series of autonomous systems, each of which creates modified versions of itself as offspring. This paper unifies the concept of computational reflection with biological systems via a new analysis of von Neumanns Universal Constructor. The result is a bio-reflective architecture that is capable of unconstrained self-modification without the problems of infinite recursion that exist in the computational counterparts. The new architecture is a blueprint for applications in Artificial Life studies, Evolutionary Algorithms, and Artificial Intelligence.

We present a subsymbolic Artificial Chemistry (ssAChem) in which all properties relevant to bonding are emergent from the underlying dynamical system (an RBN). We explore this ssAChem by evolving a seed set of atomic particles and showing the type of composite particles the system can produce.

We identify some desired mathematical properties of bonds in an Artificial Chemistry (AChem) that promote complexity and open-ended behaviour (i.e. an AChem not designed to display particular behaviours). We identify the underlying structures created by different properties of mathematical products. We use these to exploit existing algebra to generate a potentially open-ended subsymbolic Achem (ssAChem). We give examples of how our approach leads to interesting behaviour, focused on the structure of composite particles within our system.