Replicability or reproducibility? On the replication crisis in computational neuroscience and sharing only relevant detail

Replicability and reproducibility of computational models has been somewhat understudied by “the replication movement.” In this paper, we draw on methodological studies into the replicability of psychological experiments and on the mechanistic account of explanation to analyze the functions of model replications and model reproductions in computational neuroscience. We contend that model replicability, or independent researchers' ability to obtain the same output using original code and data, and model reproducibility, or independent researchers' ability to recreate a model without original code, serve different functions and fail for different reasons. This means that measures designed to improve model replicability may not enhance (and, in some cases, may actually damage) model reproducibility. We claim that although both are undesirable, low model reproducibility poses more of a threat to long-term scientific progress than low model replicability. In our opinion, low model reproducibility stems mostly from authors' omitting to provide crucial information in scientific papers and we stress that sharing all computer code and data is not a solution. Reports of computational studies should remain selective and include all and only relevant bits of code.     

Predicting patterns of long‐term adaptation and extinction with population genetics

Population genetics struggles to model extinction; standard models track the relative rather than absolute fitness of genotypes, while the exceptions describe only the short‐term transition from imminent doom to evolutionary rescue. But extinction can result from failure to adapt not only to catastrophes, but also to a backlog of environmental challenges. We model long‐term adaptation to long series of small challenges, where fitter populations reach higher population sizes. The population's long‐term fitness dynamic is well approximated by a simple stochastic Markov chain model. Long‐term persistence occurs when the rate of adaptation exceeds the rate of environmental deterioration for some genotypes. Long‐term persistence times are consistent with typical fossil species persistence times of several million years. Immediately preceding extinction, fitness declines rapidly, appearing as though a catastrophe disrupted a stably established population, even though gradual evolutionary processes are responsible. New populations go through an establishment phase where, despite being demographically viable, their extinction risk is elevated. Should the population survive long enough, extinction risk later becomes constant over time.     

The role of additive neurogenesis and synaptic plasticity in a hippocampal memory model with grid-cell like input

Recently, we presented a study of adult neurogenesis in a simplified hippocampal memory model. The network was required to encode and decode memory patterns despite changing input statistics. We showed that additive neurogenesis was a more effective adaptation strategy compared to neuronal turnover and conventional synaptic plasticity as it allowed the network to respond to changes in the input statistics while preserving representations of earlier environments. Here we extend our model to include realistic, spatially driven input firing patterns in the form of grid cells in the entorhinal cortex. We compare network performance across a sequence of spatial environments using three distinct adaptation strategies: conventional synaptic plasticity, where the network is of fixed size but the connectivity is plastic; neuronal turnover, where the network is of fixed size but units in the network may die and be replaced; and additive neurogenesis, where the network starts out with fewer initial units but grows over time. We confirm that additive neurogenesis is a superior adaptation strategy when using realistic, spatially structured input patterns. We then show that a more biologically plausible neurogenesis rule that incorporates cell death and enhanced plasticity of new granule cells has an overall performance significantly better than any one of the three individual strategies operating alone. This adaptation rule can be tailored to maximise performance of the network when operating as either a short- or long-term memory store. We also examine the time course of adult neurogenesis over the lifetime of an animal raised under different hypothetical rearing conditions. These growth profiles have several distinct features that form a theoretical prediction that could be tested experimentally. Finally, we show that place cells can emerge and refine in a realistic manner in our model as a direct result of the sparsification performed by the dentate gyrus layer.     

Robust Forward Simulations of Recurrent Hitchhiking

Evolutionary forces shape patterns of genetic diversity within populations and contribute to phenotypic variation. In particular, recurrent positive selection has attracted significant interest in both theoretical and empirical studies. However, most existing theoretical models of recurrent positive selection cannot easily incorporate realistic confounding effects such as interference between selected sites, arbitrary selection schemes, and complicated demographic processes. It is possible to quantify the effects of arbitrarily complex evolutionary models by performing forward population genetic simulations, but forward simulations can be computationally prohibitive for large population sizes (>10⁵). A common approach for overcoming these computational limitations is rescaling of the most computationally expensive parameters, especially population size. Here, we show that ad hoc approaches to parameter rescaling under the recurrent hitchhiking model do not always provide sufficiently accurate dynamics, potentially skewing patterns of diversity in simulated DNA sequences. We derive an extension of the recurrent hitchhiking model that is appropriate for strong selection in small population sizes and use it to develop a method for parameter rescaling that provides the best possible computational performance for a given error tolerance. We perform a detailed theoretical analysis of the robustness of rescaling across the parameter space. Finally, we apply our rescaling algorithms to parameters that were previously inferred for Drosophila and discuss practical considerations such as interference between selected sites.     

Asynchrony adaptation reveals neural population code for audio-visual timing

The relative timing of auditory and visual stimuli is a critical cue for determining whether sensory signals relate to a common source and for making inferences about causality. However, the way in which the brain represents temporal relationships remains poorly understood. Recent studies indicate that our perception of multisensory timing is flexible--adaptation to a regular inter-modal delay alters the point at which subsequent stimuli are judged to be simultaneous. Here, we measure the effect of audio-visual asynchrony adaptation on the perception of a wide range of sub-second temporal relationships. We find distinctive patterns of induced biases that are inconsistent with the previous explanations based on changes in perceptual latency. Instead, our results can be well accounted for by a neural population coding model in which: (i) relative audio-visual timing is represented by the distributed activity across a relatively small number of neurons tuned to different delays; (ii) the algorithm for reading out this population code is efficient, but subject to biases owing to under-sampling; and (iii) the effect of adaptation is to modify neuronal response gain. These results suggest that multisensory timing information is represented by a dedicated population code and that shifts in perceived simultaneity following asynchrony adaptation arise from analogous neural processes to well-known perceptual after-effects.     

Effects of Time-Dependent Stimuli in a Competitive Neural Network Model of Perceptual Rivalry

We analyze a competitive neural network model of perceptual rivalry that receives time-varying inputs. Time-dependence of inputs can be discrete or smooth. Spike frequency adaptation provides negative feedback that generates network oscillations when inputs are constant in time. Oscillations that resemble perceptual rivalry involve only one population being “ON” at a time, which represents the dominance of a single percept at a time. As shown in Laing and Chow (J. Comput. Neurosci. 12(1):39–53, 2002), for sufficiently high contrast, one can derive relationships between dominance times and contrast that agree with Levelt’s propositions (Levelt in On binocular rivalry, 1965). Time-dependent stimuli give rise to novel network oscillations where both, one, or neither populations are “ON” at any given time. When a single population receives an interrupted stimulus, the fundamental mode of behavior we find is phase-locking, where the temporally driven population locks its state to the stimulus. Other behaviors are analyzed as bifurcations from this forced oscillation, using fast/slow analysis that exploits the slow timescale of adaptation. When both populations receive time-varying input, we find mixtures of fusion and sole population dominance, and we partition parameter space into particular oscillation types. Finally, when a single population’s input contrast is smoothly varied in time, 1:n mode-locked states arise through period-adding bifurcations beyond phase-locking. Our results provide several testable predictions for future psychophysical experiments on perceptual rivalry.     

On the Accessibility of Adaptive Phenotypes of a Bacterial Metabolic Network

The mechanisms by which adaptive phenotypes spread within an evolving population after their emergence are understood fairly well. Much less is known about the factors that influence the evolutionary accessibility of such phenotypes, a pre-requisite for their emergence in a population. Here, we investigate the influence of environmental quality on the accessibility of adaptive phenotypes of Escherichia coli''s central metabolic network. We used an established flux-balance model of metabolism as the basis for a genotype-phenotype map (GPM). We quantified the effects of seven qualitatively different environments (corresponding to both carbohydrate and gluconeogenic metabolic substrates) on the structure of this GPM. We found that the GPM has a more rugged structure in qualitatively poorer environments, suggesting that adaptive phenotypes could be intrinsically less accessible in such environments. Nevertheless, on average ∼74% of the genotype can be altered by neutral drift, in the environment where the GPM is most rugged; this could allow evolving populations to circumvent such ruggedness. Furthermore, we found that the normalized mutual information (NMI) of genotype differences relative to phenotype differences, which measures the GPM''s capacity to transmit information about phenotype differences, is positively correlated with (simulation-based) estimates of the accessibility of adaptive phenotypes in different environments. These results are consistent with the predictions of a simple analytic theory that makes explicit the relationship between the NMI and the speed of adaptation. The results suggest an intuitive information-theoretic principle for evolutionary adaptation; adaptation could be faster in environments where the GPM has a greater capacity to transmit information about phenotype differences. More generally, our results provide insight into fundamental environment-specific differences in the accessibility of adaptive phenotypes, and they suggest opportunities for research at the interface between information theory and evolutionary biology.     

Interpreting voltage-sensitivity of gap junctions as a mechanism of cardiac memory

Memory in the nervous system is essentially a network effect, resulting from activity-dependent synaptic modification in a network of neurons. Like the nervous system, the heart is a network of cardiac cells electrically coupled by gap junctions. The heart too has memory, termed cardiac memory, whereby the effect of an external electrical activation persists long after the presentation of stimulus is terminated. We have earlier proposed that adaptation of gap junctions, as a function of membrane voltages of the cells that are coupled by the gap junctions, is related to cardiac memory [V.S. Chakravarthy, J. Ghosh, On Hebbian-like adaption in heart muscle: a proposal for "Cardiac Memory", Biol. Cybern. 76 (1997) 207, J. Krishnan, V.S. Chakravarthy, S. Radhakrishnan, On the role of gap junctions on cardiac memory effect, Comput. Cardiol. 32 (2005) 13]. Using the proposed mechanism, we demonstrate memory effect using computational models of interacting cell pairs. In this paper, we address the biological validity of the proposed mechanism of gap junctional adaptation. It is known from electrophysiology of gap junctions that the conductance of these channels adapts as a function of junctional voltage. At a first sight, this form of voltage dependence seems to be at variance with the form required by our mechanism. But we show, with the help of a theoretical model, that the proposed mechanism of voltage-dependent adaptation of gap junctions, is compatible with the known voltage-sensitivity of gap junctions observed in electrophysiological studies. Our analysis suggests a new significance of the voltage-sensitivity of gap junctions and its possible link to the phenomenon of cardiac memory.     

Analysis of avian bone response to mechanical loading, Part Two: Development of a computational connected cellular network to study bone intercellular communication

Mechanical loading-induced signals are hypothesized to be transmitted and integrated by connected bone cells before reaching the bone surfaces where adaptation occurs. A computational connected cellular network (CCCN) model is developed to explore how bone cells perceive and transmit the signals through intercellular communication. This is part two of a two-part study in which a CCCN is developed to study the intercellular communication within a grid of bone cells. The excitation signal was computed as the loading-induced bone fluid shear stress in part one. Experimentally determined bone adaptation responses (Gross et al. in J Bone Miner Res 12:982-988, 1997 and Judex et al. in J Bone Miner Res 12:1737-1745, 1997) are correlated with the fluid shear stress by the CCCN, which adjusts cell sensitivities (loading and signal thresholds) and connection weights. Intercellular communication patterns extracted by the CCCN indicate the cell population responsible for perceiving the loading-induced signal, and loading threshold is shown to play an important role in regulating the bone response.     

The evolution and evolutionary consequences of marginal thermostability in proteins

When we seek to explain the characteristics of living systems in their evolutionary context, we are often interested in understanding how and why certain properties arose through evolution, and how these properties then affected the continuing evolutionary process. This endeavor has been assisted by the use of simple computational models that have properties characteristic of natural living systems but allow simulations over evolutionary timescales with full transparency. We examine a model of the evolution of a gene under selective pressure to code for a protein that exists in a prespecified folded state at a given growth temperature. We observe the emergence of proteins with modest stabilities far below those possible with the model, with a denaturation temperature tracking the simulation temperature, despite the absence of selective pressure for such marginal stability. This demonstrates that neither observations of marginally stable proteins, nor even instances where increased stability interferes with function, provide evidence that marginal stability is an adaptation. Instead the marginal stability is the result of a balance between predominantly destabilizing mutations and selection that shifts depending on effective population size. Even if marginal stability is not an adaptation, the natural tendency of proteins toward marginal stability, and the range of stabilities that occur during evolution, may have significant effect on the evolutionary process.     

Continuous Attractor Network Model for Conjunctive Position-by-Velocity Tuning of Grid Cells

The spatial responses of many of the cells recorded in layer II of rodent medial entorhinal cortex (MEC) show a triangular grid pattern, which appears to provide an accurate population code for animal spatial position. In layer III, V and VI of the rat MEC, grid cells are also selective to head-direction and are modulated by the speed of the animal. Several putative mechanisms of grid-like maps were proposed, including attractor network dynamics, interactions with theta oscillations or single-unit mechanisms such as firing rate adaptation. In this paper, we present a new attractor network model that accounts for the conjunctive position-by-velocity selectivity of grid cells. Our network model is able to perform robust path integration even when the recurrent connections are subject to random perturbations.     

Diversification rates increase with population size and resource concentration in an unstructured habitat

Understanding the mechanisms controlling the generation and maintenance of biodiversity provides some of the planet's greatest and most pressing challenges. Variation in resource concentration, which varies widely at multiple scales, may cause biodiversity to increase, decrease, or exhibit a unimodal response and underlying mechanisms remain obscure. We established experimental cultures of long-term stationary phase (LTSP) Escherichia coli to test whether per capita heterozygosity varies with resource concentration, and, if so, whether population sizes associated with different resource concentrations contributed to these patterns. Our results provide the clearest example to date of increasing per capita heterozygosity with increasing resource concentration. Further, our experimental manipulations of population size, independent of resource concentration, provide the first unequivocal evidence that population size is one of the underlying factors controlling per capita heterozygosity along such resource gradients. Specifically, we show that cultures with higher maximum population sizes, associated with higher resource concentrations, have higher per capita heterozygosity. These experiments provide the first experimental evidence for an underappreciated factor controlling biodiversity along resource gradients--population size. This direct evidence of population size influencing diversification rates has implications for regional and global scale patterns of biodiversity.     

No accident: genetic codes freeze in error-correcting patterns of the standard genetic code

The standard genetic code poses a challenge in understanding the evolution of information processing at a fundamental level of biological organization. Genetic codes are generally coadapted with, or 'frozen' by, the protein-coding genes that they translate, and so cannot easily change by natural selection. Yet the standard code has a significantly non-random pattern that corrects common errors in the transmission of information in protein-coding genes. Because of the freezing effect and for other reasons, this pattern has been proposed not to be due to selection but rather to be incidental to other evolutionary forces or even entirely accidental. We present results from a deterministic population genetic model of code-message coevolution. We explicitly represent the freezing effect of genes on genetic codes and the perturbative effect of changes in genetic codes on genes. We incorporate characteristic patterns of mutation and translational error, namely, transition bias and positional asymmetry, respectively. Repeated selection over small successive changes produces genetic codes that are substantially, but not optimally, error correcting. In particular, our model reproduces the error-correcting patterns of the standard genetic code. Aspects of our model and results may be applicable to the general problem of adaptation to error in other natural information-processing systems.     

A general modeling framework for describing spatially structured population dynamics

Variation in movement across time and space fundamentally shapes the abundance and distribution of populations. Although a variety of approaches model structured population dynamics, they are limited to specific types of spatially structured populations and lack a unifying framework. Here, we propose a unified network‐based framework sufficiently novel in its flexibility to capture a wide variety of spatiotemporal processes including metapopulations and a range of migratory patterns. It can accommodate different kinds of age structures, forms of population growth, dispersal, nomadism and migration, and alternative life‐history strategies. Our objective was to link three general elements common to all spatially structured populations (space, time and movement) under a single mathematical framework. To do this, we adopt a network modeling approach. The spatial structure of a population is represented by a weighted and directed network. Each node and each edge has a set of attributes which vary through time. The dynamics of our network‐based population is modeled with discrete time steps. Using both theoretical and real‐world examples, we show how common elements recur across species with disparate movement strategies and how they can be combined under a unified mathematical framework. We illustrate how metapopulations, various migratory patterns, and nomadism can be represented with this modeling approach. We also apply our network‐based framework to four organisms spanning a wide range of life histories, movement patterns, and carrying capacities. General computer code to implement our framework is provided, which can be applied to almost any spatially structured population. This framework contributes to our theoretical understanding of population dynamics and has practical management applications, including understanding the impact of perturbations on population size, distribution, and movement patterns. By working within a common framework, there is less chance that comparative analyses are colored by model details rather than general principles.     

Population dynamics in migratory networks

Migratory animals are comprised of a complex series of interconnected breeding and nonbreeding populations. Because individuals in any given population can arrive from a variety of sites the previous season, predicting how different populations will respond to environmental change can be challenging. In this study, we develop a population model composed of a network of breeding and wintering sites to show how habitat loss affects patterns of connectivity and species abundance. When the costs of migration are evenly distributed, habitat loss at a single site can increase the degree of connectivity (mixing) within the entire network, which then acts to buffer global populations from declines. However, the degree to which populations are buffered depends on where habitat loss occurs within the network: a site that has the potential to receive individuals from multiple populations in the opposite season will lead to smaller declines than a site that is more isolated. In other cases when there are equal costs of migration to two or more sites in the opposite season, habitat loss can result in some populations becoming segregated (disconnected) from the rest of the network. The geographic structure of the network can have a significant influence on relative population sizes of sites in the same season and can also affect the overall degree of mixing in the network, even when sites are of equal intrinsic quality. When a migratory network is widely spaced and migration costs are high, an equivalent habitat loss will lead to a larger decline in global population size than will occur in a network where the overall costs of migration are low. Our model provides an important foundation to test predictions related to habitat loss in real-world migratory networks and demonstrates that migratory networks will likely produce different dynamics from traditional metapopulations. Our results provide strong evidence that estimating population connectivity is a prerequisite for successfully predicting changes in migratory populations.     

Local adaptation and enhanced virulence of Nosema granulosis artificially introduced into novel populations of its crustacean host, Gammarus duebeni

Local adaptation theory predicts that, on average, most parasite species should be locally adapted to their hosts (more suited to hosts from local than distant populations). Local adaptation has been studied for many horizontally transmitted parasites, however, vertically transmitted parasites have received little attention. Here we present the first study of local adaptation in an animal/parasite system where the parasite is vertically transmitted. We investigate local adaptation and patterns of virulence in a crustacean host infected with the vertically transmitted microsporidian Nosema granulosis. Nosema granulosis is vertically transmitted to successive generations of its crustacean host, Gammarus duebeni and infects up to 46% of adult females in natural populations. We investigate local adaptation using artificial horizontal infection of different host populations in the UK. Parasites were artificially inoculated from a donor population into recipient hosts from the sympatric population and into hosts from three allopatric populations in the UK. The parasite was successfully established in hosts from all populations regardless of location, infecting 45% of the recipients. Nosema granulosis was vertically (transovarially) transmitted to 39% of the offspring of artificially infected females. Parasite burden (intensity of infection) in developing embryos differed significantly between host populations and was an order of magnitude higher in the sympatric population, suggesting some degree of host population specificity with the parasite adapted to its local host population. In contrast with natural infections, artificial infection with the parasite resulted in substantial virulence, with reduced host fecundity (24%) and survival (44%) of infected hosts from all the populations regardless of location. We discuss our findings in relation to theories of local adaptation and parasite-host coevolution.     

Does reaction-diffusion support the duality of fragmentation effect?

There is a gap between single-species model predictions, and empirical studies, regarding the effect of habitat fragmentation per se, i.e., a process involving the breaking apart of habitat without loss of habitat. Empirical works indicate that fragmentation can have positive as well as negative effects, whereas, traditionally, single-species models predict a negative effect of fragmentation. Within the class of reaction-diffusion models, studies almost unanimously predict such a detrimental effect. In this paper, considering a single-species reaction-diffusion model with a removal – or similarly harvesting – term, in two dimensions, we find both positive and negative effects of fragmentation of the reserves, i.e., the protected regions where no removal occurs. Fragmented reserves lead to higher population sizes for time-constant removal terms. On the other hand, when the removal term is proportional to the population density, higher population sizes are obtained on aggregated reserves, but maximum yields are attained on fragmented configurations, and for intermediate harvesting intensities.