Cognitive Architecture with Evolutionary Dynamics Solves Insight Problem.

In this paper, we show that a neurally implemented a cognitive architecture with evolutionary dynamics can solve the four-tree problem. Our model, called Darwinian Neurodynamics, assumes that the unconscious mechanism of problem solving during insight tasks is a Darwinian process. It is based on the evolution of patterns that represent candidate solutions to a problem, and are stored and reproduced by a population of attractor networks. In our first experiment, we used human data as a benchmark and showed that the model behaves comparably to humans: it shows an improvement in performance if it is pretrained and primed appropriately, just like human participants in Kershaw et al. (2013)'s experiment. In the second experiment, we further investigated the effects of pretraining and priming in a two-by-two design and found a beginner's luck type of effect: solution rate was highest in the condition that was primed, but not pretrained with patterns relevant for the task. In the third experiment, we showed that deficits in computational capacity and learning abilities decreased the performance of the model, as expected. We conclude that Darwinian Neurodynamics is a promising model of human problem solving that deserves further investigation.

Grand Views of Evolution.

Despite major advances in evolutionary theories, some aspects of evolution remain neglected: whether evolution: would come to a halt without abiotic change; is unbounded and open-ended; or is progressive and something beyond fitness is maximized. Here, we discuss some models of ecology and evolution and argue that ecological change, resulting in Red Queen dynamics, facilitates (but does not ensure) innovation. We distinguish three forms of open-endedness. In weak open-endedness, novel phenotypes can occur indefinitely. Strong open-endedness requires the continual appearance of evolutionary novelties and/or innovations. Ultimate open-endedness entails an indefinite increase in complexity, which requires unlimited heredity. Open-ended innovation needs exaptations that generate novel niches. This can result in new traits and new rules as the dynamics unfolds, suggesting that evolution is not fully algorithmic.

Evolutionary interplay between structure, energy and epistasis in the coat protein of the ϕX174 phage family.

Viral capsids are structurally constrained by interactions among the amino acids (AAs) of their constituent proteins. Therefore, epistasis is expected to evolve among physically interacting sites and to influence the rates of substitution. To study the evolution of epistasis, we focused on the major structural protein of the ϕX174 phage family by first reconstructing the ancestral protein sequences of 18 species using a Bayesian statistical framework. The inferred ancestral reconstruction differed at eight AAs, for a total of 256 possible ancestral haplotypes. For each ancestral haplotype and the extant species, we estimated, in silico, the distribution of free energies and epistasis of the capsid structure. We found that free energy has not significantly increased but epistasis has. We decomposed epistasis up to fifth order and found that higher-order epistasis sometimes compensates pairwise interactions making the free energy seem additive. The dN/dS ratio is low, suggesting strong purifying selection, and that structure is under stabilizing selection. We synthesized phages carrying ancestral haplotypes of the coat protein gene and measured their fitness experimentally. Our findings indicate that stabilizing mutations can have higher fitness, and that fitness optima do not necessarily coincide with energy minima.

Breeding novel solutions in the brain: a model of Darwinian neurodynamics.

Background: The fact that surplus connections and neurons are pruned during development is well established. We complement this selectionist picture by a proof-of-principle model of evolutionary search in the brain, that accounts for new variations in theory space. We present a model for Darwinian evolutionary search for candidate solutions in the brain. Methods: We combine known components of the brain - recurrent neural networks (acting as attractors), the action selection loop and implicit working memory - to provide the appropriate Darwinian architecture. We employ a population of attractor networks with palimpsest memory. The action selection loop is employed with winners-share-all dynamics to select for candidate solutions that are transiently stored in implicit working memory. Results: We document two processes: selection of stored solutions and evolutionary search for novel solutions. During the replication of candidate solutions attractor networks occasionally produce recombinant patterns, increasing variation on which selection can act. Combinatorial search acts on multiplying units (activity patterns) with hereditary variation and novel variants appear due to (i) noisy recall of patterns from the attractor networks, (ii) noise during transmission of candidate solutions as messages between networks, and, (iii) spontaneously generated, untrained patterns in spurious attractors. Conclusions: Attractor dynamics of recurrent neural networks can be used to model Darwinian search. The proposed architecture can be used for fast search among stored solutions (by selection) and for evolutionary search when novel candidate solutions are generated in successive iterations. Since all the suggested components are present in advanced nervous systems, we hypothesize that the brain could implement a truly evolutionary combinatorial search system, capable of generating novel variants.

Neuronal boost to evolutionary dynamics.

Standard evolutionary dynamics is limited by the constraints of the genetic system. A central message of evolutionary neurodynamics is that evolutionary dynamics in the brain can happen in a neuronal niche in real time, despite the fact that neurons do not reproduce. We show that Hebbian learning and structural synaptic plasticity broaden the capacity for informational replication and guided variability provided a neuronally plausible mechanism of replication is in place. The synergy between learning and selection is more efficient than the equivalent search by mutation selection. We also consider asymmetric landscapes and show that the learning weights become correlated with the fitness gradient. That is, the neuronal complexes learn the local properties of the fitness landscape, resulting in the generation of variability directed towards the direction of fitness increase, as if mutations in a genetic pool were drawn such that they would increase reproductive success. Evolution might thus be more efficient within evolved brains than among organisms out in the wild.

Stability and response of polygenic traits to stabilizing selection and mutation.

When polygenic traits are under stabilizing selection, many different combinations of alleles allow close adaptation to the optimum. If alleles have equal effects, all combinations that result in the same deviation from the optimum are equivalent. Furthermore, the genetic variance that is maintained by mutation-selection balance is [Formula: see text] per locus, where µ is the mutation rate and S the strength of stabilizing selection. In reality, alleles vary in their effects, making the fitness landscape asymmetric and complicating analysis of the equilibria. We show that that the resulting genetic variance depends on the fraction of alleles near fixation, which contribute by [Formula: see text], and on the total mutational effects of alleles that are at intermediate frequency. The interplay between stabilizing selection and mutation leads to a sharp transition: alleles with effects smaller than a threshold value of [Formula: see text] remain polymorphic, whereas those with larger effects are fixed. The genetic load in equilibrium is less than for traits of equal effects, and the fitness equilibria are more similar. We find that if the optimum is displaced, alleles with effects close to the threshold value sweep first, and their rate of increase is bounded by [Formula: see text] Long-term response leads in general to well-adapted traits, unlike the case of equal effects that often end up at a suboptimal fitness peak. However, the particular peaks to which the populations converge are extremely sensitive to the initial states and to the speed of the shift of the optimum trait value.

Amino acid fermentation at the origin of the genetic code.

There is evidence that the genetic code was established prior to the existence of proteins, when metabolism was powered by ribozymes. Also, early proto-organisms had to rely on simple anaerobic bioenergetic processes. In this work I propose that amino acid fermentation powered metabolism in the RNA world, and that this was facilitated by proto-adapters, the precursors of the tRNAs. Amino acids were used as carbon sources rather than as catalytic or structural elements. In modern bacteria, amino acid fermentation is known as the Stickland reaction. This pathway involves two amino acids: the first undergoes oxidative deamination, and the second acts as an electron acceptor through reductive deamination. This redox reaction results in two keto acids that are employed to synthesise ATP via substrate-level phosphorylation. The Stickland reaction is the basic bioenergetic pathway of some bacteria of the genus Clostridium. Two other facts support Stickland fermentation in the RNA world. First, several Stickland amino acid pairs are synthesised in abiotic amino acid synthesis. This suggests that amino acids that could be used as an energy substrate were freely available. Second, anticodons that have complementary sequences often correspond to amino acids that form Stickland pairs. The main hypothesis of this paper is that pairs of complementary proto-adapters were assigned to Stickland amino acids pairs. There are signatures of this hypothesis in the genetic code. Furthermore, it is argued that the proto-adapters formed double strands that brought amino acid pairs into proximity to facilitate their mutual redox reaction, structurally constraining the anticodon pairs that are assigned to these amino acid pairs. Significance tests which randomise the code are performed to study the extent of the variability of the energetic (ATP) yield. Random assignments can lead to a substantial yield of ATP and maintain enough variability, thus selection can act and refine the assignments into a proto-code that optimises the energetic yield. Monte Carlo simulations are performed to evaluate the establishment of these simple proto-codes, based on amino acid substitutions and codon swapping. In all cases, donor amino acids are assigned to anticodons composed of U+G, and have low redundancy (1-2 codons), whereas acceptor amino acids are assigned to the the remaining codons. These bioenergetic and structural constraints allow for a metabolic role for amino acids before their co-option as catalyst cofactors.

The contribution of statistical physics to evolutionary biology.

Evolutionary biology shares many concepts with statistical physics: both deal with populations, whether of molecules or organisms, and both seek to simplify evolution in very many dimensions. Often, methodologies have undergone parallel and independent development, as with stochastic methods in population genetics. Here, we discuss aspects of population genetics that have embraced methods from physics: non-equilibrium statistical mechanics, travelling waves and Monte-Carlo methods, among others, have been used to study polygenic evolution, rates of adaptation and range expansions. These applications indicate that evolutionary biology can further benefit from interactions with other areas of statistical physics; for example, by following the distribution of paths taken by a population through time.

The statistical mechanics of a polygenic character under stabilizing selection, mutation and drift.

By exploiting an analogy between population genetics and statistical mechanics, we study the evolution of a polygenic trait under stabilizing selection, mutation and genetic drift. This requires us to track only four macroscopic variables, instead of the distribution of all the allele frequencies that influence the trait. These macroscopic variables are the expectations of: the trait mean and its square, the genetic variance, and of a measure of heterozygosity, and are derived from a generating function that is in turn derived by maximizing an entropy measure. These four macroscopics are enough to accurately describe the dynamics of the trait mean and of its genetic variance (and in principle of any other quantity). Unlike previous approaches that were based on an infinite series of moments or cumulants, which had to be truncated arbitrarily, our calculations provide a well-defined approximation procedure. We apply the framework to abrupt and gradual changes in the optimum, as well as to changes in the strength of stabilizing selection. Our approximations are surprisingly accurate, even for systems with as few as five loci. We find that when the effects of drift are included, the expected genetic variance is hardly altered by directional selection, even though it fluctuates in any particular instance. We also find hysteresis, showing that even after averaging over the microscopic variables, the macroscopic trajectories retain a memory of the underlying genetic states.

Density-dependence as a size-independent regulatory mechanism.

The growth function of populations is central in biomathematics. The main dogma is the existence of density-dependence mechanisms, which can be modelled with distinct functional forms that depend on the size of the population. One important class of regulatory functions is the theta-logistic, which generalizes the logistic equation. Using this model as a motivation, this paper introduces a simple dynamical reformulation that generalizes many growth functions. The reformulation consists of two equations, one for population size, and one for the growth rate. Furthermore, the model shows that although population is density-dependent, the dynamics of the growth rate does not depend either on population size, nor on the carrying capacity. Actually, the growth equation is uncoupled from the population size equation, and the model has only two parameters, a Malthusian parameter rho and a competition coefficient theta. Distinct sign combinations of these parameters reproduce not only the family of theta-logistics, but also the van Bertalanffy, Gompertz and Potential Growth equations, among other possibilities. It is also shown that, except for two critical points, there is a general size-scaling relation that includes those appearing in the most important allometric theories, including the recently proposed Metabolic Theory of Ecology. With this model, several issues of general interest are discussed such as the growth of animal population, extinctions, cell growth and allometry, and the effect of environment over a population.

Dynamic response of cancer under the influence of immunological activity and therapy.

The dynamical basis of tumoral growth has been controversial. Many models have been proposed to explain cancer development. The descriptions employ exponential, potential, logistic or Gompertzian growth laws. Some of these models are concerned with the interaction between cancer and the immunological system. Among other properties, these models are concerned with the microscopic behavior of tumors and the emergence of cancer. We propose a modification of a previous model by Stepanova, which describes the specific immunological response against cancer. The modification consists of the substitution of a Gompertian law for the exponential rate used for tumoral growth. This modification is motivated by the numerous works confirming that Gompertz's equation correctly describes solid tumor growth. The modified model predicts that near zero, tumors always tend to grow. Immunological contraposition never suffices to induce a complete regression of the tumor. Instead, a stable microscopic equilibrium between cancer and immunological activity can be attained. In other words, our model predicts that the theory of immune surveillance is plausible. A macroscopic equilibrium in which the system develops cancer is also possible. In this case, immunological activity is depleted. This is consistent with the phenomena of cancer tolerance. Both equilibrium points can coexist or can exist without the other. In all cases the fixed point at zero tumor size is unstable. Since immunity cannot induce a complete tumor regression, a therapy is required. We include constant-dose therapies and show that they are insufficient. Final levels of immunocompetent cells and tumoral cells are finite, thus post-treatment regrowth of the tumor is certain. We also evaluate late-intensification therapies which are successful. They induce an asymptotic regression to zero tumor size. Immune response is also suppressed by the therapy, and thus plays a negligible role in the remission. We conclude that treatment evaluation should be successful without taking into account immunological effects.

New late-intensification schedules for cancer treatments.

The nonlinear Gompertz equation that describes the evolutions of tumors under different treatments was investigated. It was shown that late logarithmic intensification asymptotically reduces the tumor cell population to zero. Using the same total amount of therapy, the schedule following a logarithmic function produces a larger reduction of tumor cell population than the schedule using a therapy with constant intensity. On the other hand, a logarithmic intensification should be more tolerable by the patients than a faster temporal growth therapy. We have shown that during the treatment the dose intensity should not be decreased at any time while the therapy is applied, because this will allow the tumor to relapse. The therapy intensity should be continuously increased as possible. We have solved an optimization problem for several late intensification treatments by the constraints of the total dose and the maximum individual (or daily) dose. Based on our results, we have designed new chemotherapy and radiotherapy treatments.