Allen Institute For Brain Science Research Articles

Cell-specific expression of key mitochondrial enzymes limits OXPHOS in astrocytes of the adult human neocortex and hippocampal formation

The astrocyte-to-neuron lactate shuttle model entails that, upon glutamatergic neurotransmission, glycolytically derived pyruvate in astrocytes is mainly converted to lactate instead of being entirely catabolized in mitochondria. The mechanism of this metabolic rewiring and its occurrence in human brain are unclear. Here by using immunohistochemistry (4 brains) and imaging mass cytometry (8 brains) we show that astrocytes of the adult human neocortex and hippocampal formation express barely detectable amounts of mitochondrial proteins critical for performing oxidative phosphorylation (OXPHOS). These data are corroborated by queries of transcriptomes (107 brains) of neuronal versus non-neuronal cells fetched from the Allen Institute for Brain Science for genes coding for a much larger repertoire of entities contributing to OXPHOS, showing that human non-neuronal elements barely expressed mRNAs coding for such proteins. With less OXPHOS, human brain astrocytes are thus bound to produce more lactate to avoid interruption of glycolysis.

Genoarchitectural Definition of the Adult Mouse Mesocortical Ring: A Contribution to Cortical Ring Theory.

Data mining was performed at the databases of the Allen Institute for Brain Science (RRID:SCR_017001) searching for genes expressed selectively throughout the adult mouse mesocortex (transitional cortex ring predicted within the concentric ring theory of mammalian cortical structure, in contrast with central isocortex [ICx] and peripheral allocortex). We aimed to explore a shared molecular profile selective of all or most mesocortex areas. This approach checks and corroborates the precision of other previous definitory criteria, such as poor myelination and high kainate receptor level. Another aim was to examine which cortical areas properly belong to mesocortex. A total of 34 positive adult selective marker genes of mesocortex were identified, jointly with 12 negative selective markers, making a total of 46 markers. All of them identify the same set of cortical areas surrounding the molecularly different ICx as well as excluding adjacent allocortex. Four representative mesocortex markers-Crym, Lypd1, Cdh13, and Smoc2-are amply illustrated, jointly with complementary material including myelin basic protein, to check myelination, and Rorb, to check layer 4 presence. The retrosplenial (ReSp) area, long held to be mesocortical, does not share any of the 46 markers of mesocortex and instead expresses Nr4a2 and Tshz2, selective parahippocampal allocortex markers. Moreover, it is not hypomyelinic and lacks a Rorb-positive layer 4, aspects generally present in mesocortex. Exclusion of the ReSp area from the mesocortex ring reveals the latter to be closed at this locus instead by two adjacent areas previously thought to be associative visual ICx (reidentified here molecularly as postsplenial and parasplenial mesocortex areas). The concepts of ICx, mesocortex, and parahippocampal allocortex are thus subtly modified by substantial molecular evidence.

Oscillating neural circuits: Phase, amplitude, and the complex normal distribution.

Multiple oscillating time series are typically analyzed in the frequency domain, where coherence is usually said to represent the magnitude of the correlation between two signals at a particular frequency. The correlation being referenced is complex-valued and is similar to the real-valued Pearson correlation in some ways but not others. We discuss the dependence among oscillating series in the context of the multivariate complex normal distribution, which plays a role for vectors of complex random variables analogous to the usual multivariate normal distribution for vectors of real-valued random variables. We emphasize special cases that are valuable for the neural data we are interested in and provide new variations on existing results. We then introduce a complex latent variable model for narrowly band-pass-filtered signals at some frequency, and show that the resulting maximum likelihood estimate produces a latent coherence that is equivalent to the magnitude of the complex canonical correlation at the given frequency. We also derive an equivalence between partial coherence and the magnitude of complex partial correlation, at a given frequency. Our theoretical framework leads to interpretable results for an interesting multivariate dataset from the Allen Institute for Brain Science.

Action Potential Parameters and Spiking Behavior of Cortical Neurons: A Statistical Analysis for Designing Spiking Neural Networks

Cortical neurons exhibit several spiking dynamics both in in-vivo and in-vitro experiments. Neural spikes or action potentials (APs) are also observed in various shapes and forms. Statistical correlation between action potential parameters and associated spiking behaviour of a neuron is discussed in this paper. Three fundamental parameters width, height and energy of an AP along with spiking frequency and inter spike interval (ISI) are extracted for 91 human cortical neurons selected from Allen Institute for Brain Science (AIBS) database. It has been shown that neurons firing narrow, short and low energy APs have higher spiking frequency compared to the neurons with wide and taller APs. For a rise in excitation, it has been presented that information gain for neurons firing wider spikes is less compared to information gain for neurons firing narrow spikes. It has been shown that neurons with low spiking frequency and high spiking frequency dissipate energy of similar order for total spiking activity for similar excitation. Implications of the statistical inferences drawn are explained for a computational model of a spiking neuron. Effect of changing AP width on the overall dynamics of a spiking neural network is also highlighted. Key findings of this study will be useful for designing spiking neural networks for various cognitive applications.

Sleepy Mice Case Study: Implementation and Assessment.

Case studies are a valuable teaching tool to engage students in course content using real-world scenarios. As part of the High-throughput Discovery Science & Inquiry-based Case Studies for Today's Students (HITS) Research Coordination Network (RCN), our team has created the Sleepy Mice Case Study for students to engage with RStudio and the Allen Institute for Brain Science's open access high-throughput sleep dataset on mice. Sleep is important for health, a familiar concern to college students, and was a basis for this case study. In this case, students completed an initial homework assignment, in-class work, and a final take-home application assignment. The case study was implemented in synchronous and asynchronous Introductory Neuroscience courses, a Biopsychology course, and a Human Anatomy and Physiology course, reflecting its versatility. The case can be used to teach course-specific learning objectives such as sleep-related content and/or science data processing skills. The case study was successful as shown by gains in student scores and confidence in achieving learning objectives. Most students reported enjoying learning about sleep deprivation course content using the case study. Best practices based on instructor experiences in implementation are also included to facilitate future use so that the Sleepy Mice Case Study can be used to teach content and/or research-related skills in various courses and modalities.

Cell-type-specific densities in mouse somatosensory cortex derived from scRNA-seq and in situ RNA hybridization.

Cells in the mammalian cerebral cortex exhibit layer-dependent patterns in their distribution. Classical methods of determining cell type distributions typically employ a painstaking process of large-scale sampling and characterization of cellular composition. We found that by combining in situ hybridization (ISH) images with cell-type-specific transcriptomes, position-dependent cortical composition in P56 mouse could be estimated in the somatosensory cortex. The method uses ISH images from the Allen Institute for Brain Science. There are two novel aspects of the methodology. First, it is not necessary to select a subset of genes that are particular for a cell type of interest, nor is it necessary to only use ISH images with low variability among samples. Second, the method also compensated for differences in soma size and incompleteness of the transcriptomes. The soma size compensation is particularly important in order to obtain quantitative estimates since relying on bulk expression alone would overestimate the contribution of larger cells. Predicted distributions of broader classes of cell types agreed with literature distributions. The primary result is that there is a high degree of substructure in the distribution of transcriptomic types beyond the resolution of layers. Furthermore, transcriptomic cell types each exhibited characteristic soma size distributions. Results suggest that the method could also be employed to assign transcriptomic cell types to well-aligned image sets in the entire brain.

Mapping of morpho-electric features to molecular identity of cortical inhibitory neurons.

Knowledge of the cell-type-specific composition of the brain is useful in order to understand the role of each cell type as part of the network. Here, we estimated the composition of the whole cortex in terms of well characterized morphological and electrophysiological inhibitory neuron types (me-types). We derived probabilistic me-type densities from an existing atlas of molecularly defined cell-type densities in the mouse cortex. We used a well-established me-type classification from rat somatosensory cortex to populate the cortex. These me-types were well characterized morphologically and electrophysiologically but they lacked molecular marker identity labels. To extrapolate this missing information, we employed an additional dataset from the Allen Institute for Brain Science containing molecular identity as well as morphological and electrophysiological data for mouse cortical neurons. We first built a latent space based on a number of comparable morphological and electrical features common to both data sources. We then identified 19 morpho-electrical clusters that merged neurons from both datasets while being molecularly homogeneous. The resulting clusters best mirror the molecular identity classification solely using available morpho-electrical features. Finally, we stochastically assigned a molecular identity to a me-type neuron based on the latent space cluster it was assigned to. The resulting mapping was used to derive inhibitory me-types densities in the cortex.

Benefits of exploring the effect of the gut microbiota on the brain

Benefits of exploring the effect of the gut microbiota on the brain Danique Mulder, Mirjam Bloemendaal & Alejandro Arias Vásquez from the departments of Psychiatry & Human Genetics, Radboud University Medical Center, look at gut microbiota and our understanding of its effects on the human brain. Dr. Christof Koch, Chief Scientist and President of the Allen Institute for Brain Science once said: “We don’t even understand the brain of a worm,” when asked how close we were to understanding how the human brain works (1). One thing we have been able to discover is that our brain does not function fractionated or in isolation. A wealth of brain studies tells us, in fact, that multiple and sometimes large areas of the brain are involved in even the most mundane memory tasks and when strong emotions are involved, millions of neurons can become more active and help us modify behaviors. Moreover, the brain is intrinsically linked with other bodily functions, including those taking place in the gastro-intestinal tract (GIt).

Constructive connectomics: How neuronal axons get from here to there using gene-expression maps derived from their family trees.

During brain development, billions of axons must navigate over multiple spatial scales to reach specific neuronal targets, and so build the processing circuits that generate the intelligent behavior of animals. However, the limited information capacity of the zygotic genome puts a strong constraint on how, and which, axonal routes can be encoded. We propose and validate a mechanism of development that can provide an efficient encoding of this global wiring task. The key principle, confirmed through simulation, is that basic constraints on mitoses of neural stem cells—that mitotic daughters have similar gene expression to their parent and do not stray far from one another—induce a global hierarchical map of nested regions, each marked by the expression profile of its common progenitor population. Thus, a traversal of the lineal hierarchy generates a systematic sequence of expression profiles that traces a staged route, which growth cones can follow to their remote targets. We have analyzed gene expression data of developing and adult mouse brains published by the Allen Institute for Brain Science, and found them consistent with our simulations: gene expression indeed partitions the brain into a global spatial hierarchy of nested contiguous regions that is stable at least from embryonic day 11.5 to postnatal day 56. We use this experimental data to demonstrate that our axonal guidance algorithm is able to robustly extend arbors over long distances to specific targets, and that these connections result in a qualitatively plausible connectome. We conclude that, paradoxically, cell division may be the key to uniting the neurons of the brain.

Multi-Data Integration Towards a Global Understanding of the Neurological Impact of Human Brain Severe Acute Respiratory Syndrome Coronavirus 2 Infection.

As the COVID-19 pandemic continues to unfold, numerous neurological symptoms emerge. The literature reports more and more manifestations of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) related to headache, dizziness, impaired consciousness, cognitive impairment, and motor disorders. Moreover, the infection of SARS-CoV-2 may have a durable neurological impact. ACE2/TMPRSS2 is the main entry point into cells for some strains of coronaviruses (CoVs), including SARS-CoV-2, which uses it to target the central nervous system (CNS). The aim of this study was to characterize the scope of the potential complex impact of a SARS-CoV-2 infection in the brain. It concerns different scales: the topographic, cognitive, sensorimotor, and genetic one. We investigated which cognitive and sensorimotor functions are associated with the brain regions where ACE2/TMPRSS2 is overexpressed, hypothesising that they might be particularly affected by the infection. Furthermore, overexpressed genes in these regions are likely to be impacted by COVID-19. This general understanding is crucial to establish the potential neurological manifestations of the infection. Data on mRNA expression levels of genes were provided by the Allen Institute for Brain Science (AIBS), and the localisation of brain functions by the LinkRbrain platform. The latter was also used to analyze the spatial overlap between ACE2/TMPRSS2 overexpression, and either function-specific brain activations or regional overexpression of other genes. The characterisation of these overexpressed genes was based on the GeneCards platform and the gene GSE164332 from the Gene Expression Omnibus database. We analysed the cognitive and sensorimotor functions whose role might be impaired, of which 88 have been categorised into seven groups: memory and recollection, motor function, pain, lucidity, emotion, sensory, and reward. Furthermore, we categorised the genes showing a significant increase in concentration of their mRNAs in the same regions where ACE2/TMPRSS2 mRNA levels are the highest. Eleven groups emerged from a bibliographical research: neurodegenerative disease, immunity, inflammation, olfactory receptor, cancer/apoptosis, executive function, senses, ischemia, motor function, myelination, and dependence. The results of this exploration could be in relation to the neurological symptoms of COVID-19. Furthermore, some genes from peripheral blood are already considered as biomarker of COVID-19. This method could generate new hypotheses to explore the neurological manifestations of COVID-19.

Enhancing Student Research Experiences with Open Data from the Allen Brain Map.

The Allen Brain Map is the main data repository for the Allen Institute for Brain Science, containing big, open datasets commonly used in neuroscience research (Allen Institute for Brain Science, 2022). Open data from the Allen Brain Map can be used to teach core concepts in neuroscience, data analysis methods, and other critical skills and knowledge to neuroscience students. These datasets can be used as the main data source for completely online lab experiences, or analyzed in combination with data students collect themselves. Applications may range in scope and format from a short worksheet used in a single class session to a coding tutorial to a guided independent research project. While open online data cannot fully replace lab experiences for learning techniques, they can be used to expose students to analysis of big data, introduce resources widely used in the field, and teach skills like statistics and coding. This article reviews potential assignment formats where big and open data can be applied, introduces selected popular resources and sample use cases for each, and discusses benefits and limitations of open online data for lab experiences. Some specific applications in the context of distance learning are also detailed.

Cortical Thickness Differences Are Associated With Chemical Synaptic Transmission Upregulated Genes in Degeneration of Mild Cognitive Impairment

Mild cognitive impairment (MCI) is a transition between normal cognition (NC) and Alzheimer’s disease (AD). Differences in cortical thickness (ΔCT) have been reported in cases that degenerate from MCI to AD. The aspects of genetic and transcriptional variation related to ΔCT are vague. In this study, using an 8-year longitudinal follow-up outcome, we investigated the genetic correlates of ΔCT in MCI subjects with degeneration from MCI to AD (MCI_AD). We employed partial least squares regression (PLSR) on brain T1-weighted magnetic resonance imaging (MRI) images of 180 participants [143 stable MCI (MCI_S) participants and 37 MCI_AD participants] and brain gene expression data from the Allen Institute for Brain Science (AIBS) database to investigate genes associated with ΔCT. We found that upregulated PLS component 1 ΔCT-related genes were enriched in chemical synaptic transmission. To verify the robustness and specificity of the results, we conducted PLSR analysis invalidation and specificity datasets and performed weighted gene co-expression network analysis instead of PLSR for the above three datasets. We also used gene expression data in the brain prefrontal cortex from the Gene Expression Omnibus (GEO) database to indirectly validate the robustness and specificity of our results. We conclude that transcriptionally upregulated genes involved in chemical synaptic transmission are strongly related to global ΔCT in MCI patients who experience degeneration from MCI to AD.

Mapping the corticoreticular pathway from cortex-wide anterograde axonal tracing in the mouse.

The corticoreticular pathway (CRP) has been implicated as an important mediator of motor recovery and rehabilitation after central nervous system damage. However, its origins, trajectory and laterality are not well understood. This study mapped the mouse CRP in comparison with the corticospinal tract (CST). We systematically searched the Allen Mouse Brain Connectivity Atlas (© 2011 Allen Institute for Brain Science) for experiments that used anterograde tracer injections into the right isocortex in mice. For each eligible experiment (N=607), CRP and CST projection strength were quantified by the tracer volume reaching the reticular formation motor nuclei (RFmotor ) and pyramids, respectively. Tracer density in each brain voxel was also correlated with RFmotor versus pyramids projection strength to explore the relative trajectories of the CRP and CST. We found significant CRP projections originating from the primary and secondary motor cortices, anterior cingulate, primary somatosensory cortex, and medial prefrontal cortex. Compared with the CST, the CRP had stronger projections from each region except the primary somatosensory cortex. Ipsilateral projections were stronger than contralateral for both tracts (above the pyramidal decussation), but the CRP projected more bilaterally than the CST. The estimated CRP trajectory was anteromedial to the CST in the internal capsule and dorsal to the CST in the brainstem. Our findings reveal a widespread distribution of CRP origins and confirm strong bilateral CRP projections, theoretically increasing the potential for partial sparing after brain lesions and contralesional compensation after unilateral injury.

Transitions in development - an interview with Tom Nowakowski.

Tom Nowakowski is an Assistant Professor at University of California San Francisco (UCSF), where he uses single-cell sequencing technologies to study neurodevelopment. He is also a Chan Zuckerberg Biohub Investigator and a Next Generation Leader at the Allen Institute for Brain Science. We met with Tom over Zoom to hear more about his career, his transition to becoming a group leader and his plans for the future.

Timing and localization of myasthenia gravis-related gene expression.

Myasthenia gravis (MG) is an acquired autoimmune disorder caused by autoantibodies binding acetylcholine receptors (AChR), muscle‐specific kinase (MuSK), agrin or low‐density lipoprotein receptor‐related protein 4 (Lrp4). These autoantibodies inhibit neuromuscular transmission by blocking the function of these proteins and thereby cause fluctuating skeletal muscle weakness. Several reports suggest that these autoantibodies might also affect the central nervous system (CNS) in MG patients. A comprehensive overview of the timing and localization of the expression of MG‐related antigens in other organs is currently lacking. To investigate the spatio‐temporal expression of MG‐related genes outside skeletal muscle, we used in silico tools to assess public expression databases. Acetylcholine esterase, nicotinic AChR α1 subunit, agrin, collagen Q, downstream of kinase‐7, Lrp4, MuSK and rapsyn were included as MG‐related genes because of their well‐known involvement in either congenital or autoimmune MG. We investigated expression of MG‐related genes in (1) all human tissues using GTEx data, (2) specific brain regions, (3) neurodevelopmental stages, and (4) cell types using datasets from the Allen Institute for Brain Sciences. MG‐related genes show heterogenous spatio‐temporal expression patterns in the human body as well as in the CNS. For each of these genes, several (new) tissues, brain areas and cortical cell types with (relatively) high expression were identified suggesting a potential role for these genes outside skeletal muscle. The possible presence of MG‐related antigens outside skeletal muscle suggests that autoimmune MG, congenital MG or treatments targeting the same proteins may affect MG‐related protein function in other organs.

Differential Spatiotemporal Expression of Type I and Type II Cadherins Associated With the Segmentation of the Central Nervous System and Formation of Brain Nuclei in the Developing Mouse.

Type I and type II classical cadherins comprise a family of cell adhesion molecules that regulate cell sorting and tissue separation by forming specific homo and heterophilic bonds. Factors that affect cadherin-mediated cell-cell adhesion include cadherin binding affinity and expression level. This study examines the expression pattern of type I cadherins (Cdh1, Cdh2, Cdh3, and Cdh4), type II cadherins (Cdh6, Cdh7, Cdh8, Cdh9, Cdh10, Cdh11, Cdh12, Cdh18, Cdh20, and Cdh24), and the atypical cadherin 13 (Cdh13) during distinct morphogenetic events in the developing mouse central nervous system from embryonic day 11.5 to postnatal day 56. Cadherin mRNA expression levels obtained from in situ hybridization experiments carried out at the Allen Institute for Brain Science (https://alleninstitute.org/) were retrieved from the Allen Developing Mouse Brain Atlas. Cdh2 is the most abundantly expressed type I cadherin throughout development, while Cdh1, Cdh3, and Cdh4 are expressed at low levels. Type II cadherins show a dynamic pattern of expression that varies between neuroanatomical structures and developmental ages. Atypical Cdh13 expression pattern correlates with Cdh2 in abundancy and localization. Analyses of cadherin-mediated relative adhesion estimated from their expression level and binding affinity show substantial differences in adhesive properties between regions of the neural tube associated with the segmentation along the anterior–posterior axis. Differences in relative adhesion were also observed between brain nuclei in the developing subpallium (basal ganglia), suggesting that differential cell adhesion contributes to the segregation of neuronal pools. In the adult cerebral cortex, type II cadherins Cdh6, Cdh8, Cdh10, and Cdh12 are abundant in intermediate layers, while Cdh11 shows a gradated expression from the deeper layer 6 to the superficial layer 1, and Cdh9, Cdh18, and Cdh24 are more abundant in the deeper layers. Person’s correlation analyses of cadherins mRNA expression patterns between areas and layers of the cerebral cortex and the nuclei of the subpallium show significant correlations between certain cortical areas and the basal ganglia. The study shows that differential cadherin expression and cadherin-mediated adhesion are associated with a wide range of morphogenetic events in the developing central nervous system including the organization of neurons into layers, the segregation of neurons into nuclei, and the formation of neuronal circuits.

Flexible annotation atlas of the mouse brain: combining and dividing brain structures of the Allen Brain Atlas while maintaining anatomical hierarchy

A brain atlas is necessary for analyzing structure and function in neuroimaging research. Although various annotation volumes (AVs) for the mouse brain have been proposed, it is common in magnetic resonance imaging (MRI) of the mouse brain that regions-of-interest (ROIs) for brain structures (nodes) are created arbitrarily according to each researcher’s necessity, leading to inconsistent ROIs among studies. One reason for such a situation is the fact that earlier AVs were fixed, i.e. combination and division of nodes were not implemented. This report presents a pipeline for constructing a flexible annotation atlas (FAA) of the mouse brain by leveraging public resources of the Allen Institute for Brain Science on brain structure, gene expression, and axonal projection. A mere two-step procedure with user-specified, text-based information and Python codes constructs FAA with nodes which can be combined or divided objectively while maintaining anatomical hierarchy of brain structures. Four FAAs with total node count of 4, 101, 866, and 1381 were demonstrated. Unique characteristics of FAA realized analysis of resting-state functional connectivity (FC) across the anatomical hierarchy and among cortical layers, which were thin but large brain structures. FAA can improve the consistency of whole brain ROI definition among laboratories by fulfilling various requests from researchers with its flexibility and reproducibility.

Uncovering Statistical Links Between Gene Expression and Structural Connectivity Patterns in the Mouse Brain

Finding links between genes and structural connectivity is of the utmost importance for unravelling the underlying mechanism of the brain connectome. In this study we identify links between the gene expression and the axonal projection density in the mouse brain, by applying a modified version of the Linked ICA method to volumetric data from the Allen Institute for Brain Science for identifying independent sources of information that link both modalities at the voxel level. We performed separate analyses on sets of projections from the visual cortex, the caudoputamen and the midbrain reticular nucleus, and we determined those brain areas, injections and genes that were most involved in independent components that link both gene expression and projection density data, while we validated their biological context through enrichment analysis. We identified representative and literature-validated cortico-midbrain and cortico-striatal projections, whose gene subsets were enriched with annotations for neuronal and synaptic function and related developmental and metabolic processes. The results were highly reproducible when including all available projections, as well as consistent with factorisations obtained using the Dictionary Learning and Sparse Coding technique. Hence, Linked ICA yielded reproducible independent components that were preserved under increasing data variance. Taken together, we have developed and validated a novel paradigm for linking gene expression and structural projection patterns in the mouse mesoconnectome, which can power future studies aiming to relate genes to brain function.

Decoding Mouse Behavior from neural activity in primary visual cortex

Background and Hypothesis: 
 Brain-machine interfaces (BMIs) aim to restore motor control to patients. To develop effective BMIs, it is imperative that we build highly accurate models relating cortical neural activity with natural movements. Neurons in the visual cortex of the brain respond strongly to visual signals, but are also modulated by fluctuating internal states such as alertness, which closely correlates pupil size. We thus hypothesized that an animal’s pupil size can be decoded directly from the activity of a population of neurons in mouse primary visual cortex. 
 
 Project Methods: 
 We used an open-access dataset from the Allen Institute for Brain Science, which consisted of in-vivo 2 photon calcium imaging data of 128 neurons recorded from primary visual cortex of an awake mouse viewing a video screen. Video imaging was used to track the mouse’s pupil size. 
 
 To relate neural activity with pupil size, we used linear regression, first smoothing the data with a Savitzky Gall filter. The model was fit on the first 80% of the data, and tested on the remaining 20%, using the computed coefficients to predict pupil size from neural activity. A Pearson correlation coefficient between the true and predicted data was computed. All analyses were performed in Python. 
 
 Results: 
 We found that there was a strong linear relationship between neural activity and pupil size, as the correlation coefficient between the true and predicted data was 0.376. Therefore, we have shown quantitatively that pupil size can be modeled based on neuronal activity within the visual cortex with a high level of accuracy. 
 
 Conclusion and Potential Impact: 
 This experiment demonstrates the feasibility of linear regression for modeling neuronal behavior in the visual system, particularly related to pupil size. Results suggest opportunities for further experimental analysis of the relationship between neural activity and other physiological markers, such as running speed.

BayeSuites: An open web framework for massive Bayesian networks focused on neuroscience

BayeSuites is the first web framework for learning, visualizing, and interpreting Bayesian networks (BNs) that can scale to tens of thousands of nodes while providing fast and friendly user experience. All the necessary features that enable this are reviewed in this paper; these features include scalability, extensibility, interoperability, ease of use, and interpretability. Scalability is the key factor in learning and processing massive networks within reasonable time; for a maintainable software open to new functionalities, extensibility and interoperability are necessary. Ease of use and interpretability are fundamental aspects of model interpretation, fairly similar to the case of the recent explainable artificial intelligence trend. We present the capabilities of our proposed framework by highlighting a real example of a BN learned from genomic data obtained from Allen Institute for Brain Science. The extensibility properties of the software are also demonstrated with the help of our BN-based probabilistic clustering implementation, together with another genomic-data example.