Evolution of Associative Learning in Chemical Networks

Abstract

Organisms that can learn about their environment and modify their behaviour appropriately during their lifetime are more likely to survive and reproduce than organisms that do not. While associative learning – the ability to detect correlated features of the environment – has been studied extensively in nervous systems, where the underlying mechanisms are reasonably well understood, mechanisms within single cells that could allow associative learning have received little attention. Here, using in silico evolution of chemical networks, we show that there exists a diversity of remarkably simple and plausible chemical solutions to the associative learning problem, the simplest of which uses only one core chemical reaction. We then asked to what extent a linear combination of chemical concentrations in the network could approximate the ideal Bayesian posterior of an environment given the stimulus history so far? This Bayesian analysis revealed the ‘memory traces’ of the chemical network. The implication of this paper is that there is little reason to believe that a lack of suitable phenotypic variation would prevent associative learning from evolving in cell signalling, metabolic, gene regulatory, or a mixture of these networks in cells.

The Evolution of Phenotypic Correlations and “Developmental Memory”

“…The phenotype of an individual at developmental time step, t, is described by a set of N phenotypic characters or traits, naturally 1 Other work has investigated the potential to implement associative learning mechanisms in various nonneural systems (e.g., metabolic networks- Fernando et al 2008); and investigated the ability of evolution to find such mechanisms (McGregor et al 2012) that can then operate within the lifetime of the individual. Here we do not select for an associative learning mechanism, we simply evolve developmental interactions.…”

“…); and investigated the ability of evolution to find such mechanisms (McGregor et al. ) that can then operate within the lifetime of the individual. Here we do not select for an associative learning mechanism, we simply evolve developmental interactions.…”

Viewpoint: From Responsive to Adaptive and Interactive Materials and Materials Systems: A Roadmap

“…The classic example of the Pavlovian dog, which salivates initially only when food is provided (unconditioned stimulus), but learns to salivate also later when it hears the steps of the assistant bringing the food or when a bell is rung prior to food provision (i.e., an initially neutral stimulus turns into a conditioned stimulus), demonstrates that signal processing routes need to become intertwined and altered due to repeated and time‐correlated signal exposure. Simplistic processes in this direction have been proposed theoretically and realizations might not be too far, at least for systems operating in the realm of synthetic biology . One of the most critical aspects in associative learning is the fact that the signal processing routes have to be truly interconnected with a rerouting of the signal processing by installation of a memory element, and that it shows important differences to simple conditioning (i.e., programming; Figure e).…”

Design of a biochemical circuit motif for learning linear functions

“…In other related theoretical work, Kim et al proposed implementations of neural networks using in vitro transcriptional circuits [51], including constructs for Hopfield networks and winner-takes-all learning systems, and circuits have been previously proposed for Hebbian learning in synthetic biological transcription networks [5,6]. In some impressive experimental work, Qian et al [8] demonstrated a biochemical implementation of a Hopfield neural network [52] using strand displacement-based 'seesaw' gates [37].…”