A study of transfer of information in animal collectives using deep learning tools

We studied how the interactions among animals in a collective allow for the transfer of information. We performed laboratory experiments to study how zebrafish in a collective follow a subset of trained animals that move towards a light when it turns on because they expect food at that location. We built some deep learning tools to distinguish from video which are the trained and the naïve animals and to detect when each animal reacts to the light turning on. These tools gave us the data to build a model of interactions that we designed to have a balance between transparency and accuracy. The model finds a low-dimensional function that describes how a naïve animal weights neighbours depending on focal and neighbour variables. According to this low-dimensional function, neighbour speed plays an important role in the interactions. Specifically, a naïve animal weights more a neighbour in front than to the sides or behind, and more so the faster the neighbour is moving; and if the neighbour moves fast enough, the differences coming from the neighbour's relative position largely disappear. From the lens of decision-making, neighbour speed acts as confidence measure about where to go. This article is part of a discussion meeting issue 'Collective behaviour through time'.

Dealing with dimensionality: the application of machine learning to multi-omics data

MOTIVATION

Machine learning (ML) methods are motivated by the need to automate information extraction from large datasets in order to support human users in data-driven tasks. This is an attractive approach for integrative joint analysis of vast amounts of omics data produced in next generation sequencing and other -omics assays. A systematic assessment of the current literature can help to identify key trends and potential gaps in methodology and applications. We surveyed the literature on ML multi-omic data integration and quantitatively explored the goals, techniques and data involved in this field. We were particularly interested in examining how researchers use ML to deal with the volume and complexity of these datasets.

RESULTS

Our main finding is that the methods used are those that address the challenges of datasets with few samples and many features. Dimensionality reduction methods are used to reduce the feature count alongside models that can also appropriately handle relatively few samples. Popular techniques include autoencoders, random forests and support vector machines. We also found that the field is heavily influenced by the use of The Cancer Genome Atlas dataset, which is accessible and contains many diverse experiments.

AVAILABILITY AND IMPLEMENTATION

All data and processing scripts are available at this GitLab repository: https://gitlab.com/polavieja_lab/ml_multi-omics_review/ or in Zenodo: https://doi.org/10.5281/zenodo.7361807.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Early-Life Social Experience Shapes Social Avoidance Reactions in Larval Zebrafish

Social experiences greatly define subsequent social behavior. Lack of such experiences, especially during critical phases of development, can severely impede the ability to behave adequately in social contexts. To date, it is not well characterized how early-life social isolation leads to social deficits and impacts development. In many model species, it is challenging to fully control social experiences, because they depend on parental care. Moreover, complex social behaviors involve multiple sensory modalities, contexts, and actions. Hence, when studying social isolation effects, it is important to parse apart social deficits from general developmental effects, such as abnormal motor learning. Here, we characterized how social experiences during early development of zebrafish larvae modulate their social behavior at 1 week of age, when social avoidance reactions can be measured as discrete swim events. We show that raising larvae in social isolation leads to enhanced social avoidance, in terms of the distance at which larvae react to one another and the strength of swim movement they use. Specifically, larvae raised in isolation use a high-acceleration escape swim, the short latency C-start, more frequently during social interactions. These behavioral differences are absent in non-social contexts. By ablating the lateral line and presenting the fish with local water vibrations, we show that lateral line inputs are both necessary and sufficient to drive enhanced social avoidance reactions. Taken together, our results show that social experience during development is a critical factor in shaping mechanosensory avoidance reactions in larval zebrafish.

Deep attention networks reveal the rules of collective motion in zebrafish

A variety of simple models has been proposed to understand the collective motion of animals. These models can be insightful but may lack important elements necessary to predict the motion of each individual in the collective. Adding more detail increases predictability but can make models too complex to be insightful. Here we report that deep attention networks can obtain a model of collective behavior that is simultaneously predictive and insightful thanks to an organization in modules. When using simulated trajectories, the model recovers the ground-truth interaction rule used to generate them, as well as the number of interacting neighbours. For experimental trajectories of large groups of 60-100 zebrafish, Danio rerio, the model obtains that interactions between pairs can approximately be described as repulsive, attractive or as alignment, but only when moving slowly. At high velocities, interactions correspond only to alignment or alignment mixed with repulsion at close distances. The model also shows that each zebrafish decides where to move by aggregating information from the group as a weighted average over neighbours. Weights are higher for neighbours that are close, in a collision path or moving faster in frontal and lateral locations. The network also extracts that the number of interacting individuals is dynamical and typically in the range 8–22, with 1–10 more important ones. Our results suggest that each animal decides by dynamically selecting information from the collective.

Simple models have traditionally been very successful, because they usually provide more insight than complicated models. This is particularly true in physics, where simple models can often give highly precise quantitative predictions. However, biology is fundamentally complex and thus it is difficult to find simple models that give precise predictions. To create models that are both precise and insightful, we propose to harness the power of deep neural networks but to confine them into modules with a low number of inputs and outputs. We trained one such model to predict the future turning side of a fish in a collective. By plotting the different modules we obtain insight about how fish interact and how they aggregate information from different neighbours. This aggregation is dynamical and shows that fish can interact with approximately 20 neighbours but can also focus on fewer neighbours, down to 1-2, when some move at higher speed in front or to the sides, are very close or are in a collision path.

Adolescents show collective intelligence which can be driven by a geometric mean rule of thumb

How effective groups are in making decisions is a long-standing question in studying human and animal behaviour. Despite the limited social and cognitive abilities of younger people, skills which are often required for collective intelligence, studies of group performance have been limited to adults. Using a simple task of estimating the number of sweets in jars, we show in two experiments that adolescents at least as young as 11 years old improve their estimation accuracy after a period of group discussion, demonstrating collective intelligence. Although this effect was robust to the overall distribution of initial estimates, when the task generated positively skewed estimates, the geometric mean of initial estimates gave the best fit to the data compared to other tested aggregation rules. A geometric mean heuristic in consensus decision making is also likely to apply to adults, as it provides a robust and well-performing rule for aggregating different opinions. The geometric mean rule is likely to be based on an intuitive logarithmic-like number representation, and our study suggests that this mental number scaling may be beneficial in collective decisions.

Species diversity rises exponentially with the number of available resources in a multi-trait competition model

Theoretical studies of ecosystem models have generally concluded that large numbers of species will not stably coexist if the species are all competing for the same limited set of resources. Here, we describe a simple multi-trait model of competition where the presence of N resources will lead to the stable coexistence of up to 2 N species. Our model also predicts that the long-term dynamics of the population will lie on a neutral attractor hyperplane. When the population shifts within the hyperplane, its dynamics will behave neutrally, while shifts which occur perpendicular to the hyperplane will be subject to restoring forces. This provides a potential explanation of why complex ecosystems might exhibit both niche-like and neutral responses to perturbations. Like the neutral theory of biodiversity, our model generates good fits to species abundance distributions in several datasets but does so without needing to evoke inter-generational stochastic effects, continuous species creation or immigration dynamics. Additionally, our model is able to explain species abundance correlations between independent but similar ecosystems separated by more than 1400 km inside the Amazonian forests.

Zebrafish aggression on the sub-second time scale: evidence for mutual motor coordination and multi-functional attack manoeuvres

Most animals fight by repeating complex stereotypic behaviours, yet the internal structure of these behaviours has rarely been dissected in detail. We characterized the internal structure of fighting behaviours by developing a machine learning pipeline that measures and classifies the behaviour of individual unmarked animals on a sub-second time scale. This allowed us to quantify several previously hidden features of zebrafish fighting strategies. We found strong correlations between the velocity of the attacker and the defender, indicating a dynamic matching of approach and avoidance efforts. While velocity matching was ubiquitous, the spatial dynamics of attacks showed phase-specific differences. Contest-phase attacks were characterized by a paradoxical sideways attraction of the retreating animal towards the attacker, suggesting that the defender combines avoidance manoeuvres with display-like manoeuvres. Post-resolution attacks lacked display-like features and the defender was avoidance focused. From the perspective of the winner, game-theory modelling further suggested that highly energetically costly post-resolution attacks occurred because the winner was trying to increase its relative dominance over the loser. Overall, the rich structure of zebrafish motor coordination during fighting indicates a greater complexity and layering of strategies than has previously been recognized.

Adversity magnifies the importance of social information in decision-making

Decision-making theories explain animal behaviour, including human behaviour, as a response to estimations about the environment. In the case of collective behaviour, they have given quantitative predictions of how animals follow the majority option. However, they have so far failed to explain that in some species and contexts social cohesion increases when conditions become more adverse (i.e. individuals choose the majority option with higher probability when the estimated quality of all available options decreases). We have found that this failure is due to modelling simplifications that aided analysis, like low levels of stochasticity or the assumption that only one choice is the correct one. We provide a more general but simple geometric framework to describe optimal or suboptimal decisions in collectives that gives insight into three different mechanisms behind this effect. The three mechanisms have in common that the private information acts as a gain factor to social information: a decrease in the privately estimated quality of all available options increases the impact of social information, even when social information itself remains unchanged. This increase in the importance of social information makes it more likely that agents will follow the majority option. We show that these results quantitatively explain collective behaviour in fish and experiments of social influence in humans.

Histone H4 acetylation regulates behavioral inter-individual variability in zebrafish

BACKGROUND

Animals can show very different behaviors even in isogenic populations, but the underlying mechanisms to generate this variability remain elusive. We use the zebrafish (Danio rerio) as a model to test the influence of histone modifications on behavior.

RESULTS

We find that laboratory and isogenic zebrafish larvae show consistent individual behaviors when swimming freely in identical wells or in reaction to stimuli. This behavioral inter-individual variability is reduced when we impair the histone deacetylation pathway. Individuals with high levels of histone H4 acetylation, and specifically H4K12, behave similarly to the average of the population, but those with low levels deviate from it. More precisely, we find a set of genomic regions whose histone H4 acetylation is reduced with the distance between the individual and the average population behavior. We find evidence that this modulation depends on a complex of Yin-yang 1 (YY1) and histone deacetylase 1 (HDAC1) that binds to and deacetylates these regions. These changes are not only maintained at the transcriptional level but also amplified, as most target regions are located near genes encoding transcription factors.

CONCLUSIONS

We suggest that stochasticity in the histone deacetylation pathway participates in the generation of genetic-independent behavioral inter-individual variability.

Microsaccadic sampling of moving image information provides Drosophila hyperacute vision

Small fly eyes should not see fine image details. Because flies exhibit saccadic visual behaviors and their compound eyes have relatively few ommatidia (sampling points), their photoreceptors would be expected to generate blurry and coarse retinal images of the world. Here we demonstrate that Drosophila see the world far better than predicted from the classic theories. By using electrophysiological, optical and behavioral assays, we found that R1-R6 photoreceptors’ encoding capacity in time is maximized to fast high-contrast bursts, which resemble their light input during saccadic behaviors. Whilst over space, R1-R6s resolve moving objects at saccadic speeds beyond the predicted motion-blur-limit. Our results show how refractory phototransduction and rapid photomechanical photoreceptor contractions jointly sharpen retinal images of moving objects in space-time, enabling hyperacute vision, and explain how such microsaccadic information sampling exceeds the compound eyes’ optical limits. These discoveries elucidate how acuity depends upon photoreceptor function and eye movements.

Fruit flies have five eyes: two large compound eyes which support vision, plus three smaller single lens eyes which are used for navigation. Each compound eye monitors 180° of space and consists of roughly 750 units, each containing eight light-sensitive cells called photoreceptors. This relatively wide spacing of photoreceptors is thought to limit the sharpness, or acuity, of vision in fruit flies. The area of the human retina (the light-sensitive surface at back of our eyes) that generates our sharpest vision contains photoreceptors that are 500 times more densely packed.

Despite their differing designs, human and fruit fly eyes work via the same general principles. If we, or a fruit fly, were to hold our gaze completely steady, the world would gradually fade from view as the eye adapted to the unchanging visual stimulus. To ensure this does not happen, animals continuously make rapid, automatic eye movements called microsaccades. These refresh the image on the retina and prevent it from fading. Yet it is not known why do they not also cause blurred vision.

Standard accounts of vision assume that the retina and the brain perform most of the information processing required, with photoreceptors simply detecting how much light enters the eye. However, Juusola, Dau, Song et al. now challenge this idea by showing that photoreceptors are specially adapted to detect the fluctuating patterns of light that enter the eye as a result of microsaccades. Moreover, fruit fly eyes resolve small moving objects far better than would be predicted based on the spacing of their photoreceptors.

The discovery that photoreceptors are well adapted to deal with eye movements changes our understanding of insect vision. The findings also disprove the 100-year-old dogma that the spacing of photoreceptors limits the sharpness of vision in compound eyes. Further studies are required to determine whether photoreceptors in the retinas of other animals, including humans, have similar properties.

Signatures of optimal control in pairs of schooling zebrafish

Animals moving in groups coordinate their motion to remain cohesive. A large amount of data and analysis of movement coordination has been obtained in several species, but we are lacking theoretical frameworks that can derive the form of coordination rules. Here, we examine whether optimal control theory can predict the rules underlying social interactions from first principles. We find that a control rule which is designed to minimize the time it would take a pair of schooling fish to form a cohesively moving unit correctly predicts the characteristics of social interactions in fish. Our methodology explains why social attraction is negatively modulated by self-motion velocity and positively modulated by partner motion velocity, and how the biomechanics of fish swimming can shape the form of social forces. Crucially, the values of all parameters in our model can be estimated from independent experiments that need not relate to measurement of social interactions. We test our theory by showing a good match with experimentally observed social interaction rules in zebrafish. In addition to providing a theoretical rationale for observed decision rules, we suggest that this framework opens new questions about tuning problems and learnability of collective behaviours.

Toward a Molecular Profile of Self-Representation

Feeling embodiment over our body or body part has a major role in the understanding of the self and control of self-actions. Even though it is crucial in our daily life, embodiment is not an homogenous phenotype across population, as quantified by implicit and explicit measures (i.e., neuroimaging or self-reports). Studies have shown differences in neuropathological conditions compared to healthy controls, but also across healthy individuals. We discuss examples of self-perception differences, and the molecular origin of embodiment, focusing on clinical cases, during the first and second section. We then discuss two important questions in this molecular-to-embodiment relationship: (i) which are the molecular levels (and their associated techniques) that can be relevant to embodiment, and (ii) which are the most adequate experiments to correlate molecular profiles and embodiment quantification across individuals. Potential answers for both questions will be outlined during the third and fourth sections, respectively, in order to design a framework to study the molecular profile of body embodiment.

The relationship between behavior acquisition and persistence abilities: Involvement of adult hippocampal neurogenesis

The influence of the learning process on the persistence of the newly acquired behavior is relevant both for our knowledge of the learning/memory mechanisms and for the educational policy. However, it is unclear whether during an operant conditioning process with a continuous reinforcement paradigm, individual differences in acquisition are also associated to differences in persistence of the acquired behavior. In parallel, adult neurogenesis has been implicated in spatial learning and memory, but the specific role of the immature neurons born in the adult brain is not well known for this process. We have addressed both questions by analyzing the relationship between water maze task acquisition scores, the persistence of the acquired behavior, and the size of the different subpopulations of immature neurons in the adult murine hippocampus. We have found that task acquisition and persistence rates were negatively correlated: the faster the animals find the water maze platform at the end of acquisition stage, the less they persist in searching for it at the learned position in a subsequent non-reinforced trial; accordingly, the correlation in the number of some new neurons' subpopulations and the acquisition rate is negative while with persistence in acquired behavior is positive. These findings reveal an unexpected relationship between the efficiency to learn a task and the persistence of the new behavior after a non-reinforcement paradigm, and suggest that the immature neurons might be involved in different roles in acquisition and persistence/extinction of a learning task. © 2016 Wiley Periodicals, Inc.

Improving Collective Estimations Using Resistance to Social Influence

Groups can make precise collective estimations in cases like the weight of an object or the number of items in a volume. However, in others tasks, for example those requiring memory or mental calculation, subjects often give estimations with large deviations from factual values. Allowing members of the group to communicate their estimations has the additional perverse effect of shifting individual estimations even closer to the biased collective estimation. Here we show that this negative effect of social interactions can be turned into a method to improve collective estimations. We first obtained a statistical model of how humans change their estimation when receiving the estimates made by other individuals. We confirmed using existing experimental data its prediction that individuals use the weighted geometric mean of private and social estimations. We then used this result and the fact that each individual uses a different value of the social weight to devise a method that extracts the subgroups resisting social influence. We found that these subgroups of individuals resisting social influence can make very large improvements in group estimations. This is in contrast to methods using the confidence that each individual declares, for which we find no improvement in group estimations. Also, our proposed method does not need to use historical data to weight individuals by performance. These results show the benefits of using the individual characteristics of the members in a group to better extract collective wisdom.

Automated image-based tracking and its application in ecology

The behavior of individuals determines the strength and outcome of ecological interactions, which drive population, community, and ecosystem organization. Bio-logging, such as telemetry and animal-borne imaging, provides essential individual viewpoints, tracks, and life histories, but requires capture of individuals and is often impractical to scale. Recent developments in automated image-based tracking offers opportunities to remotely quantify and understand individual behavior at scales and resolutions not previously possible, providing an essential supplement to other tracking methodologies in ecology. Automated image-based tracking should continue to advance the field of ecology by enabling better understanding of the linkages between individual and higher-level ecological processes, via high-throughput quantitative analysis of complex ecological patterns and processes across scales, including analysis of environmental drivers.

The ontogeny of sleep-wake cycles in zebrafish: a comparison to humans

Zebrafish (Danio rerio) are used extensively in sleep research; both to further understanding of sleep in general and also as a model of human sleep. To date, sleep studies have been performed in larval and adult zebrafish but no efforts have been made to document the ontogeny of zebrafish sleep-wake cycles. Because sleep differs across phylogeny and ontogeny it is important to validate the use of zebrafish in elucidating the neural substrates of sleep. Here we describe the development of sleep and wake across the zebrafish lifespan and how it compares to humans. We find power-law distributions to best fit wake bout data but demonstrate that exponential distributions, previously used to describe sleep bout distributions, fail to adequately account for the data in either species. Regardless, the data reveal remarkable similarities in the ontogeny of sleep cycles in zebrafish and humans. Moreover, as seen in other organisms, zebrafish sleep levels are highest early in ontogeny and sleep and wake bouts gradually consolidate to form the adult sleep pattern. Finally, sleep percentage, bout duration, bout number, and sleep fragmentation are shown to allow for meaningful comparisons between zebrafish and human sleep.

The first mecp2-null zebrafish model shows altered motor behaviors

Rett syndrome (RTT) is an X-linked neurodevelopmental disorder and one of the most common causes of mental retardation in affected girls. Other symptoms include a rapid regression of motor and cognitive skills after an apparently early normal development. Sporadic mutations in the transcription factor MECP2 has been shown to be present in more than 90% of the patients and several models of MeCP2-deficient mice have been created to understand the role of this gene. These models have pointed toward alterations in the maintenance of the central nervous system rather than its development, in line with the late onset of the disease in humans. However, the exact functions of MeCP2 remain difficult to delineate and the animal models have yielded contradictory results. Here, we present the first mecp2-null allele mutation zebrafish model. Surprisingly and in contrast to MeCP2-null mouse models, mecp2-null zebrafish are viable and fertile. They present nonetheless clear behavioral alterations during their early development, including spontaneous and sensory-evoked motor anomalies, as well as defective thigmotaxis.

Wiring economy and volume exclusion determine neuronal placement in the Drosophila brain

Wiring economy has successfully explained the individual placement of neurons in simple nervous systems like that of Caenorhabditis elegans [1-3] and the locations of coarser structures like cortical areas in complex vertebrate brains [4]. However, it remains unclear whether wiring economy can explain the placement of individual neurons in brains larger than that of C. elegans. Indeed, given the greater number of neuronal interconnections in larger brains, simply minimizing the length of connections results in unrealistic configurations, with multiple neurons occupying the same position in space. Avoiding such configurations, or volume exclusion, repels neurons from each other, thus counteracting wiring economy. Here we test whether wiring economy together with volume exclusion can explain the placement of neurons in a module of the Drosophila melanogaster brain known as lamina cartridge [5-13]. We used newly developed techniques for semiautomated reconstruction from serial electron microscopy (EM) [14] to obtain the shapes of neurons, the location of synapses, and the resultant synaptic connectivity. We show that wiring length minimization and volume exclusion together can explain the structure of the lamina microcircuit. Therefore, even in brains larger than that of C. elegans, at least for some circuits, optimization can play an important role in individual neuron placement.

The origin of behavioral bursts in decision-making circuitry

From ants to humans, the timing of many animal behaviors comes in bursts of activity separated by long periods of inactivity. Recently, mathematical modeling has shown that simple algorithms of priority-driven behavioral choice can result in bursty behavior. To experimentally test this link between decision-making circuitry and bursty dynamics, we have turned to Drosophila melanogaster. We have found that the statistics of intervals between activity periods in endogenous activity-rest switches of wild-type Drosophila are very well described by the Weibull distribution, a common distribution of bursty dynamics in complex systems. The bursty dynamics of wild-type Drosophila walking activity are shown to be determined by this inter-event distribution alone and not by memory effects, thus resembling human dynamics. Further, using mutant flies that disrupt dopaminergic signaling or the mushroom body, circuitry implicated in decision-making, we show that the degree of behavioral burstiness can be modified. These results are thus consistent with the proposed link between decision-making circuitry and bursty dynamics, and highlight the importance of using simple experimental systems to test general theoretical models of behavior. The findings further suggest that analysis of bursts could prove useful for the study and evaluation of decision-making circuitry.

Network adaptation improves temporal representation of naturalistic stimuli in Drosophila eye: II mechanisms

Retinal networks must adapt constantly to best present the ever changing visual world to the brain. Here we test the hypothesis that adaptation is a result of different mechanisms at several synaptic connections within the network. In a companion paper (Part I), we showed that adaptation in the photoreceptors (R1-R6) and large monopolar cells (LMC) of the Drosophila eye improves sensitivity to under-represented signals in seconds by enhancing both the amplitude and frequency distribution of LMCs' voltage responses to repeated naturalistic contrast series. In this paper, we show that such adaptation needs both the light-mediated conductance and feedback-mediated synaptic conductance. A faulty feedforward pathway in histamine receptor mutant flies speeds up the LMC output, mimicking extreme light adaptation. A faulty feedback pathway from L2 LMCs to photoreceptors slows down the LMC output, mimicking dark adaptation. These results underline the importance of network adaptation for efficient coding, and as a mechanism for selectively regulating the size and speed of signals in neurons. We suggest that concert action of many different mechanisms and neural connections are responsible for adaptation to visual stimuli. Further, our results demonstrate the need for detailed circuit reconstructions like that of the Drosophila lamina, to understand how networks process information.

Network adaptation improves temporal representation of naturalistic stimuli in Drosophila eye: I dynamics

Because of the limited processing capacity of eyes, retinal networks must adapt constantly to best present the ever changing visual world to the brain. However, we still know little about how adaptation in retinal networks shapes neural encoding of changing information. To study this question, we recorded voltage responses from photoreceptors (R1–R6) and their output neurons (LMCs) in the Drosophila eye to repeated patterns of contrast values, collected from natural scenes. By analyzing the continuous photoreceptor-to-LMC transformations of these graded-potential neurons, we show that the efficiency of coding is dynamically improved by adaptation. In particular, adaptation enhances both the frequency and amplitude distribution of LMC output by improving sensitivity to under-represented signals within seconds. Moreover, the signal-to-noise ratio of LMC output increases in the same time scale. We suggest that these coding properties can be used to study network adaptation using the genetic tools in Drosophila, as shown in a companion paper (Part II).

Optimally wired subnetwork determines neuroanatomy of Caenorhabditis elegans

Wiring cost minimization has successfully explained many structures of nervous systems. However, in the nematode Caenorhabditis elegans, for which anatomical data are most detailed, wiring economy is thought to play only a partial role and alone has failed to account for the grouping of neurons into ganglia [Chen BL, Hall DH, Chklovskii DB (2006) Proc Natl Acad Sci USA 103:4723-4728; Kaiser M, Hilgetag CC (2006) PLoS Comput Biol 2:e95; Ahn Y-Y, Jeong H, Kim BJ (2006) Physica A 367:531-537]. Here, we test the hypothesis that optimally wired subnetworks can exist within nonoptimal networks, thus allowing wiring economy to give an improved prediction of spatial structure. We show in C. elegans that the small subnetwork of wires connecting sensory and motor neurons with sensors and muscles, comprising only 15% of connections, is close to optimal and alone predicts the main features of the spatial segregation of neurons into ganglia and encephalization. Moreover, a method to dissect networks into optimal and nonoptimal components is shown to find a large near-optimal subnetwork of 84% of neurons with a very low position error of 5.4%, and that explains clustering of neurons into ganglia and encephalization to fine detail. In general, we expect realistic networks not to be globally optimal in wire cost. We thus propose the strategy of using near-optimal subnetworks to understand neuroanatomical structure.

Sodium pumps adapt spike bursting to stimulus statistics

Pump activity is a homeostatic mechanism that maintains ionic gradients. Here we examined whether the slow reduction in excitability induced by sodium-pump activity that has been seen in many neuronal types is also involved in neuronal coding. We took intracellular recordings from a spike-bursting sensory neuron in the leech Hirudo medicinalis in response to naturalistic tactile stimuli with different statistical distributions. We show that regulation of excitability by sodium pumps is necessary for the neuron to make different responses depending on the statistical context of the stimuli. In particular, sodium-pump activity allowed spike-burst sizes and rates to code not for stimulus values per se, but for their ratio with the standard deviation of the stimulus distribution. Modeling further showed that sodium pumps can be a general mechanism of adaptation to statistics on the time scale of 1 min. These results implicate the ubiquitous pump activity in the adaptation of neural codes to statistics.

Coding with spike shapes and graded potentials in cortical networks

In cortical neurones, analogue dendritic potentials are thought to be encoded into patterns of digital spikes. According to this view, neuronal codes and computations are based on the temporal patterns of spikes: spike times, bursts or spike rates. Recently, we proposed an 'action potential waveform code' for cortical pyramidal neurones in which the spike shape carries information. Broader somatic action potentials are reliably produced in response to higher conductance input, allowing for four times more information transfer than spike times alone. This information is preserved during synaptic integration in a single neurone, as back-propagating action potentials of diverse shapes differentially shunt incoming postsynaptic potentials and so participate in the next round of spike generation. An open question has been whether the information in action potential waveforms can also survive axonal conduction and directly influence synaptic transmission to neighbouring neurones. Several new findings have now brought new light to this subject, showing cortical information processing that transcends the classical models.

Age-independent synaptogenesis by phosphoinositide 3 kinase

Synapses are specialized communication points between neurons, and their number is a major determinant of cognitive abilities. These dynamic structures undergo developmental- and activity-dependent changes. During brain aging and certain diseases, synapses are gradually lost, causing mental decline. It is, thus, critical to identify the molecular mechanisms controlling synapse number. We show here that the levels of phosphoinositide 3 kinase (PI3K) regulate synapse number in both Drosophila larval motor neurons and adult brain projection neurons. The supernumerary synapses induced by PI3K overexpression are functional and elicit changes in behavior. Remarkably, PI3K activation induces synaptogenesis in aged adult neurons as well. We demonstrate that persistent PI3K activity is necessary for synapse maintenance. We also report that PI3K controls the expression and localization of synaptic markers in human neuroblastoma cells, suggesting that PI3K synaptogenic activity is conserved in humans. Thus, we propose that PI3K stimulation can be applied to prevent or delay synapse loss in normal aging and in neurological disorders.

Neuronal algorithms that detect the temporal order of events

One of the basic operations in sensory processing is the computation of the temporal order of excitation of sensors. Motivated by the discrepancy between models and experiments at high signal contrast, we obtain families of algorithms by solutions of a general set of equations that define temporal order detection as an input-to-output relationship. Delays and nonlinear operations are the basis of all algorithms found, but different algorithmic structures exist when the operations are multiplications, OR gates, different types of AND-NOT logical gates, or concatenated AND-NOT gates. Among others, we obtain the Hassenstein-Reichardt model, a network using a multiplicative operation that has been proposed to explain fly optomotor behavior. We also find extensions of the Barlow-Levick model (based on an AND-NOT gate with delayed inhibition and nondelayed excitation as inputs), originally proposed to explain the bipolar cell response of the rabbit retina to motion stimuli. In the extended models, there are two more steps, another AND-NOT gate, and a subtraction or two subtractions that make the model responsive only to motion. In response to low-contrast inputs, the concatenated AND-NOT gates or the AND-NOT gate followed by a subtraction in these new models act as the multiplicative operation in the Hassenstein-Reichardt model. At high contrast, the new models behave like the Hassenstein-Reichardt model except that they are independent of contrast as observed experimentally.

Feedback network controls photoreceptor output at the layer of first visual synapses in Drosophila

At the layer of first visual synapses, information from photoreceptors is processed and transmitted towards the brain. In fly compound eye, output from photoreceptors (R1-R6) that share the same visual field is pooled and transmitted via histaminergic synapses to two classes of interneuron, large monopolar cells (LMCs) and amacrine cells (ACs). The interneurons also feed back to photoreceptor terminals via numerous ligand-gated synapses, yet the significance of these connections has remained a mystery. We investigated the role of feedback synapses by comparing intracellular responses of photoreceptors and LMCs in wild-type Drosophila and in synaptic mutants, to light and current pulses and to naturalistic light stimuli. The recordings were further subjected to rigorous statistical and information-theoretical analysis. We show that the feedback synapses form a negative feedback loop that controls the speed and amplitude of photoreceptor responses and hence the quality of the transmitted signals. These results highlight the benefits of feedback synapses for neural information processing, and suggest that similar coding strategies could be used in other nervous systems.

Variability in noise-driven integrator neurons

Neural variability in the presence of noise has been studied mainly in resonator neurons, such as Hodgkin-Huxley or FitzHugh-Nagumo models. Here we investigate this variability for integrator neurons, whose excitability is due to a saddle-node bifurcation of the rest state instead of a Hopf bifurcation. Using simple theoretical expressions for the interspike times distributions, we obtain coefficients of variation in good agreement with numerical calculations in realistic neuron models. The main features of this coefficient as a function of noise depend on the refractory period and on the presence of bistability. The bistability is responsible for the existence of two different time scales in the spiking behavior giving an antiresonance effect.

Reliable biological communication with realistic constraints

Communication in biological systems must deal with noise and metabolic or temporal constraints. We include these constraints into information theory to obtain the distributions of signal usage corresponding to a maximal rate of information transfer given any noise structure and any constraints. Generalized versions of the Boltzmann, Gaussian, or Poisson distributions are obtained for linear, quadratic and temporal constraints, respectively. These distributions are shown to imply that biological transformations must dedicate a larger output range to the more probable inputs and less to the outputs with higher noise and higher participation in the constraint. To show the general theory of reliable communication at work, we apply these results to biochemical and neuronal signaling. Noncooperative enzyme kinetics is shown to be suited for transfer of a high signal quality when the input distribution has a maximum at low concentrations while cooperative kinetics for near-Gaussian input statistics. Neuronal codes based on spike rates, spike times or bursts have to balance signal quality and cost-efficiency and at the network level imply sparseness and uncorrelation within the limits of noise, cost, and processing operations.

The rate of information transfer of naturalistic stimulation by graded potentials

We present a method to measure the rate of information transfer for any continuous signals of finite duration without assumptions. After testing the method with simulated responses, we measure the encoding performance of Calliphora photoreceptors. We find that especially for naturalistic stimulation the responses are nonlinear and noise is nonadditive, and show that adaptation mechanisms affect signal and noise differentially depending on the time scale, structure, and speed of the stimulus. Different signaling strategies for short- and long-term and dim and bright light are found for this graded system when stimulated with naturalistic light changes.

Errors drive the evolution of biological signalling to costly codes

Reduction of costs in biological signalling seems an evolutionary advantage, but recent experiments have shown signalling codes shifted to signals of high cost with an underutilization of low-cost signals. Here I derive a theory for efficient signalling that includes both errors and costs as constraints and I show that errors in the efficient translation of biological states into signals can shift codes to higher costs, effectively performing a quality control. The statistical structure of signal usage is predicted to be of a generalized Boltzmann form that penalizes signals that are costly and sensitive to errors. This predicted distribution of signal usage against signal cost has two main features: an exponential tail required for cost efficiency and an underutilization of the low-cost signals required to protect the signalling quality from the errors. These predictions are shown to correspond quantitatively to the experiments in which gathering signal statistics is feasible as in visual cortex neurons.