Chemical product design refers to the practice of developing novel chemical products given properties to be optimised and constraints to be satisfied. Strategies for chemical product design can be based on multi-objective constrained optimisation in a large search space of compounds whose properties are uncertain and partially known. Advances in machine learning, multi-objective optimisation, formal representation of chemical compounds and identified correlations between molecular structures and relevant properties, have fostered increased interest in computer-based techniques to identify candidate compounds for innovation in chemical products. In this paper we empirically explore a combination of state-of-the-art machine learning and evolutionary multi-objective optimisation methods to support chemical product design. In order to ground our arguments as concrete examples, we consider the design of domestic detergents, and explore how automating computational design can be controlled via specification of hyper-parameters, so as to generate solutions (detergents) with desired features. Our results contribute to the methodological problem of automating chemical product design, and more broadly functional molecular design.

Recent work has demonstrated the viability of DNA robotics and artificial molecular machines for molecular transportation and cargo sorting with potential applications in manufacturing responsive molecular devices, programmable therapeutics, and autonomous chemical synthesis. We extend previous work on cooperative molecular transportation using artificial molecular machines, where we similarly functionalize DNA-conjugated microtubules driven by kinesin motor proteins. DNA-functionalized microtubules propelled by surface-adhered kinesin motors enable the self-organization of molecular swarms, where such swarms load and transport cargo (microbead) in a simulated chemical environment. We demonstrate programmable molecular swarms for cargo sorting and cooperative transport. Cargo loading occurs when sufficient microtubules are at the same location as the cargo, and cargo unloading occurs at specific points in the environment through interaction with localized DNA species. Our contribution is the design of a chemotaxis molecular controller, forcing the swarm to tumble (random change direction) when the system is not following a molecular gradient corresponding to the cargo type, thus directing it to specific points for cargo unloading. This work thus contributes to the open problem of how to best design programmable molecular machines for various tasks in microscopic environments.

The drug discovery process broadly follows the sequence of high-throughput screening, optimisation, synthesis, testing, and finally, clinical trials. We investigate methods for accelerating this process with machine learning algorithms that can automatically design novel ligands for biological targets. Recent work has demonstrated the viability of deep reinforcement learning, generative adversarial networks and auto-encoders. Here, we extend state-of-the-art deep reinforcement learning molecular modification algorithms and, through the integration of molecular docking simulations, apply them to automatically design novel antagonists for the adenosine triphosphate binding site of Plasmodium falciparum phosphatidylinositol 4-kinase, an enzyme essential to the malaria parasite’s development within an infected host. We demonstrated that such an algorithm was capable of designing novel molecular graphs with better DSs than the best DSs in a set of reference molecules. There reference set here was a set of 1,011 structural analogues of napthyridine, imidazopyridazine, and aminopyradine.

Extended Abstract Many researchers hypothesize that language adaptation, as with other evolutionary processes, entails both directed selection and random drift (Sapir, 1921; McMahon, 1994; Croft, 2000; Baxter et al., 2006; Van de Velde, 2014; Steels and Szathmáry, 2018). However, the specific contributions of these processes to language evolution remains an open question. It is well established that language evolution is not necessarily driven by selection, for example, speakers preferring specific word variants (Andersen, 1987; Blythe, 2012; Hamilton et al., 2016; Newberry et al., 2017). Extending related work (Kandler et al., 2017), we use computational agent-based models to elucidate the impact of individual-level bias (speaker prestige) on population-level dynamics (average word similarity), where word diversity is measured by Levenshtein similarity (Levenshtein, 1966). Agents interacted in iterative language games (Kirby et al., 2014), to name and thus converse about resource types (A, B). Such object types represented conversation topics (Karjus et al., 2020c), where resource value indicated agent bias for conversing about (evolving words for) popular topics. For a null model comparison, we comparatively evaluated random drift versus directed word evolution on evolving word similarity, where using directed evolution, agent bias for adopting specific words (about resource types) increased with speaker agent social prestige (fitness). While previous work has demonstrated selective advantages of various forms of speaker sociolinguistic prestige including competing word variants and borrowed words (Abrams and Strogatz, 2003; Labov, 2011; J. Hernández-Campoy and J. Conde-Silvestre, 2012; Kauhanen, 2017; Calude et al., 2017; Monaghan and Roberts, 2019; Karjus et al., 2020a,c), there has been little research on the impact of speaker prestige on word diversity in language evolution.

The social brain hypothesis posits that the evolution of big brains (neural complexity) in groups of social organisms is the evolutionary result of cognitive challenges associated with various complex interactions and the need to process and solve complex social tasks. This study aims to investigate the environmental and evolutionary conditions under which neural complexity evolves without sacrificing collective behavioral efficacy. Using an evolutionary collective robotics system this research evaluates the impact of imposing a fitness cost on evolving increased neural complexity in robot groups that must operate (accomplish cooperative tasks) in environments of varying complexity. Results indicate that for all environments tested, imposing a cost on neural complexity induces the evolution of smaller neural controllers that are comparably effective to more complex controllers.

It has been argued that much of evolution takes place in the absence of fitness gradients. Such periods of evolution can be analysed by examining the mutational network formed by sequences of equal fitness, that is, the neutral network. It has been demonstrated that, in large populations under a high mutation rate, the population distribution over the neutral network and average mutational robustness are given by the principal eigenvector and eigenvalue, respectively, of the network’s adjacency matrix. However, little progress has been made towards understanding the manner in which the topology of the neutral network influences the resulting population distribution and robustness. In this work, we use numerical methods and network models to enhance our understanding of how populations distribute themselves over neutral networks. We demonstrate that, in the presence of certain topological features, the population will undergo an exploration catastrophe and become confined to a small portion of the network. These results provide insight into the behaviour of populations on neutral networks, demonstrating that neutrality does not necessarily lead to an exploration of genotype/phenotype space or an associated increase in population diversity.

This study reports work in progress on an Agent-Based Model (ABM) that critically explores specific theories that have gained prominence in the study of Egyptian state formation in the last two decades. The goal is to develop a model that simulates the evolution of complex social and economic networks by incorporating idiosyncrasies of human character and decision making, in order to create more plausible historical reconstructions. This study’s ABM focuses on wealth accumulation and loss in a simple agrarian society within an environment that simulates the Upper Egyptian landscape (ca. 4000 BC), the time period when clear evidence of economic and social inequality among Upper Egyptian households can be observed in the archaeological record. Understanding the establishment of permanent, entrenched inequality is crucial for exploring the development of social complexity and hierarchy.

Previous research has demonstrated that computational models of Gene Regulatory Networks (GRNs) can adapt so as to increase their evolvability, where evolvability is defined as a populations responsiveness to environmental change. In such previous work, phenotypes have been represented as bit strings formed by concatenating the activations of the GRN after simulation. This research is an extension where previous results supporting the evolvability of GRNs are replicated, however, the phenotype space is enriched with time and space dynamics with an evolutionary robotics task environment. It was found that a GRN encoding used in the evolution of a way-point navigation behavior in a fluctuating environment results in (robot controller) populations becoming significantly more responsive (evolvable) over time. This is as compared to a direct encoding of controllers which was unable to improve its evolvability in the same task environment.

Robustness and evolvability have traditionally been seen as conflicting properties of evolutionary systems, due to the fact that selection requires heritable variation on which to operate. Various recent studies have demonstrated that organisms evolving in environments fluctuating non-randomly become better at adapting to these fluctuations, that is, increase their evolvability. It has been suggested that this is due to the emergence of biases in the mutational neighborhoods of genotypes. This paper examines a potential consequence of these observations, that a large bias in certain areas of genotype space will lead to increased robustness in corresponding phenotypes. The evolution of boolean networks, which bear similarity to models of gene regulatory networks, is simulated in environments which fluctuate between task targets. It was found that an increase in evolvability is concomitant with the emergence of highly robust genotypes, where evolvability was defined as the populations adaptability. Analysis of the genotype space elucidated that evolution finds regions containing robust genotypes coding for one of the target phenotypes, where these regions overlap or are situated in close proximity. Results indicate that genotype space topology impacts the relationship between robustness and evolvability, where the separation of robust regions coding for the various targets was detrimental to evolvability.