Brain-wide Electrical Spatiotemporal Dynamics Encode Depression Vulnerability.

Brain-wide fluctuations in local field potential oscillations reflect emergent network-level signals that mediate behavior. Cracking the code whereby these oscillations coordinate in time and space (spatiotemporal dynamics) to represent complex behaviors would provide fundamental insights into how the brain signals emotional pathology. Using machine learning, we discover a spatiotemporal dynamic network that predicts the emergence of major depressive disorder (MDD)-related behavioral dysfunction in mice subjected to chronic social defeat stress. Activity patterns in this network originate in prefrontal cortex and ventral striatum, relay through amygdala and ventral tegmental area, and converge in ventral hippocampus. This network is increased by acute threat, and it is also enhanced in three independent models of MDD vulnerability. Finally, we demonstrate that this vulnerability network is biologically distinct from the networks that encode dysfunction after stress. Thus, these findings reveal a convergent mechanism through which MDD vulnerability is mediated in the brain.

Geometric deep learning enables 3D kinematic profiling across species and environments.

Comprehensive descriptions of animal behavior require precise three-dimensional (3D) measurements of whole-body movements. Although two-dimensional approaches can track visible landmarks in restrictive environments, performance drops in freely moving animals, due to occlusions and appearance changes. Therefore, we designed DANNCE to robustly track anatomical landmarks in 3D across species and behaviors. DANNCE uses projective geometry to construct inputs to a convolutional neural network that leverages learned 3D geometric reasoning. We trained and benchmarked DANNCE using a dataset of nearly seven million frames that relates color videos and rodent 3D poses. In rats and mice, DANNCE robustly tracked dozens of landmarks on the head, trunk, and limbs of freely moving animals in naturalistic settings. We extended DANNCE to datasets from rat pups, marmosets, and chickadees, and demonstrate quantitative profiling of behavioral lineage during development.

Supervising the Decoder of Variational Autoencoders to Improve Scientific Utility.

Probabilistic generative models are attractive for scientific modeling because their inferred parameters can be used to generate hypotheses and design experiments. This requires that the learned model provides an accurate representation of the input data and yields a latent space that effectively predicts outcomes relevant to the scientific question. Supervised Variational Autoencoders (SVAEs) have previously been used for this purpose, as a carefully designed decoder can be used as an interpretable generative model of the data, while the supervised objective ensures a predictive latent representation. Unfortunately, the supervised objective forces the encoder to learn a biased approximation to the generative posterior distribution, which renders the generative parameters unreliable when used in scientific models. This issue has remained undetected as reconstruction losses commonly used to evaluate model performance do not detect bias in the encoder. We address this previously-unreported issue by developing a second-order supervision framework (SOS-VAE) that updates the decoder parameters, rather than the encoder, to induce a predictive latent representation. This ensures that the encoder maintains a reliable posterior approximation and the decoder parameters can be effectively interpreted. We extend this technique to allow the user to trade-off the bias in the generative parameters for improved predictive performance, acting as an intermediate option between SVAEs and our new SOS-VAE. We also use this methodology to address missing data issues that often arise when combining recordings from multiple scientific experiments. We demonstrate the effectiveness of these developments using synthetic data and electrophysiological recordings with an emphasis on how our learned representations can be used to design scientific experiments.

Electome network factors: Capturing emotional brain networks related to health and disease.

Therapeutic development for mental disorders has been slow despite the high worldwide prevalence of illness. Unfortunately, cellular and circuit insights into disease etiology have largely failed to generalize across individuals that carry the same diagnosis, reflecting an unmet need to identify convergent mechanisms that would facilitate optimal treatment. Here, we discuss how mesoscale networks can encode affect and other cognitive processes. These networks can be discovered through electrical functional connectome (electome) analysis, a method built upon explainable machine learning models for analyzing and interpreting mesoscale brain-wide signals in a behavioral context. We also outline best practices for identifying these generalizable, interpretable, and biologically relevant networks. Looking forward, translational electome analysis can span species and various moods, cognitive processes, or other brain states, supporting translational medicine. Thus, we argue that electome analysis provides potential translational biomarkers for developing next-generation therapeutics that exhibit high efficacy across heterogeneous disorders.

Predicting emerging chemical content in consumer products using machine learning.

Chemical ingredients in consumer products are continually changing. To understand our exposure to chemicals and their consequent risk, we need to know their concentrations in products, or chemical weight fractions. Unfortunately, manufacturers rarely report comprehensive weight fraction data on product labels. The goal of this study was to evaluate the utility of machine learning strategies for predicting weight fractions when chemical constituent data are limited. A "data-poor" framework was developed and tested using a small dataset on consumer products containing engineered nanomaterials to represent emerging substances. A second, more traditional framework was applied to a "data-rich" product dataset comprised of bulk-scale organic chemicals for comparison purposes. Feature variables included chemical properties, functional use categories (e.g., antimicrobial), product categories (e.g., makeup), product matrix categories, and whether weight fractions were manufacturer-reported or experimentally obtained. Classification into three weight fraction bins was done using a random forest or nonlinear support vector classifier. An ablation study revealed that functional use data improved predictive performance when included alongside chemical property data, suggesting the utility of functional use categories in evaluating the safety and sustainability of emerging chemicals. Models could roughly stratify material-product observations into order of magnitude weight fractions with moderate success; the best of these achieved an average balanced accuracy of 73% on the nanomaterials product data. Framework comparisons also revealed a positive trend in sample size versus average balanced accuracy, suggesting great promise for machine learning approaches with continued investment in chemical data collection.

Novel trends of genome evolution in highly complex tropical sponge microbiomes.

BACKGROUND: Tropical members of the sponge genus Ircinia possess highly complex microbiomes that perform a broad spectrum of chemical processes that influence host fitness. Despite the pervasive role of microbiomes in Ircinia biology, it is still unknown how they remain in stable association across tropical species. To address this question, we performed a comparative analysis of the microbiomes of 11 Ircinia species using whole-metagenomic shotgun sequencing data to investigate three aspects of bacterial symbiont genomes-the redundancy in metabolic pathways across taxa, the evolution of genes involved in pathogenesis, and the nature of selection acting on genes relevant to secondary metabolism. RESULTS: A total of 424 new, high-quality bacterial metagenome-assembled genomes (MAGs) were produced for 10 Caribbean Ircinia species, which were evaluated alongside 113 publicly available MAGs sourced from the Pacific species Ircinia ramosa. Evidence of redundancy was discovered in that the core genes of several primary metabolic pathways could be found in the genomes of multiple bacterial taxa. Across hosts, the metagenomes were depleted in genes relevant to pathogenicity and enriched in eukaryotic-like proteins (ELPs) that likely mimic the hosts' molecular patterning. Finally, clusters of steroid biosynthesis genes (CSGs), which appear to be under purifying selection and undergo horizontal gene transfer, were found to be a defining feature of Ircinia metagenomes. CONCLUSIONS: These results illustrate patterns of genome evolution within highly complex microbiomes that illuminate how associations with hosts are maintained. The metabolic redundancy within the microbiomes could help buffer the hosts from changes in the ambient chemical and physical regimes and from fluctuations in the population sizes of the individual microbial strains that make up the microbiome. Additionally, the enrichment of ELPs and depletion of LPS and cellular motility genes provide a model for how alternative strategies to virulence can evolve in microbiomes undergoing mixed-mode transmission that do not ultimately result in higher levels of damage (i.e., pathogenicity) to the host. Our last set of results provides evidence that sterol biosynthesis in Ircinia-associated bacteria is widespread and that these molecules are important for the survival of bacteria in highly complex Ircinia microbiomes. Video Abstract.

Prenatal environmental stressors impair postnatal microglia function and adult behavior in males.

Gestational exposure to environmental toxins and socioeconomic stressors is epidemiologically linked to neurodevelopmental disorders with strong male bias, such as autism. We model these prenatal risk factors in mice by co-exposing pregnant dams to an environmental pollutant and limited-resource stress, which robustly activates the maternal immune system. Only male offspring display long-lasting behavioral abnormalities and alterations in the activity of brain networks encoding social interactions. Cellularly, prenatal stressors diminish microglial function within the anterior cingulate cortex, a central node of the social coding network, in males during early postnatal development. Precise inhibition of microglial phagocytosis within the anterior cingulate cortex (ACC) of wild-type (WT) mice during the same critical period mimics the impact of prenatal stressors on a male-specific behavior, indicating that environmental stressors alter neural circuit formation in males via impairing microglia function during development.

Brain-wide electrical dynamics encode individual appetitive social behavior.

The architecture whereby activity across many brain regions integrates to encode individual appetitive social behavior remains unknown. Here we measure electrical activity from eight brain regions as mice engage in a social preference assay. We then use machine learning to discover a network that encodes the extent to which individual mice engage another mouse. This network is organized by theta oscillations leading from prelimbic cortex and amygdala that converge on the ventral tegmental area. Network activity is synchronized with cellular firing, and frequency-specific activation of a circuit within this network increases social behavior. Finally, the network generalizes, on a mouse-by-mouse basis, to encode individual differences in social behavior in healthy animals but fails to encode individual behavior in a 'high confidence' genetic model of autism. Thus, our findings reveal the architecture whereby the brain integrates distributed activity across timescales to encode an appetitive brain state underlying individual differences in social behavior.

The Relationship Between Microbiomes and Selective Regimes in the Sponge Genus Ircinia.

Sponges are often densely populated by microbes that benefit their hosts through nutrition and bioactive secondary metabolites; however, sponges must simultaneously contend with the toxicity of microbes and thwart microbial overgrowth. Despite these fundamental tenets of sponge biology, the patterns of selection in the host sponges' genomes that underlie tolerance and control of their microbiomes are still poorly understood. To elucidate these patterns of selection, we performed a population genetic analysis on multiple species of Ircinia from Belize, Florida, and Panama using an F ST -outlier approach on transcriptome-annotated RADseq loci. As part of the analysis, we delimited species boundaries among seven growth forms of Ircinia. Our analyses identified balancing selection in immunity genes that have implications for the hosts' tolerance of high densities of microbes. Additionally, our results support the hypothesis that each of the seven growth forms constitutes a distinct Ircinia species that is characterized by a unique microbiome. These results illuminate the evolutionary pathways that promote stable associations between host sponges and their microbiomes, and that potentially facilitate ecological divergence among Ircinia species.

On Target Shift in Adversarial Domain Adaptation.

Discrepancy between training and testing domains is a fundamental problem in the generalization of machine learning techniques. Recently, several approaches have been proposed to learn domain invariant feature representations through adversarial deep learning. However, label shift, where the percentage of data in each class is different between domains, has received less attention. Label shift naturally arises in many contexts, especially in behavioral studies where the behaviors are freely chosen. In this work, we propose a method called Domain Adversarial nets for Target Shift (DATS) to address label shift while learning a domain invariant representation. This is accomplished by using distribution matching to estimate label proportions in a blind test set. We extend this framework to handle multiple domains by developing a scheme to upweight source domains most similar to the target domain. Empirical results show that this framework performs well under large label shift in synthetic and real experiments, demonstrating the practical importance.

Extracting Relationships by Multi-Domain Matching.

In many biological and medical contexts, we construct a large labeled corpus by aggregating many sources to use in target prediction tasks. Unfortunately, many of the sources may be irrelevant to our target task, so ignoring the structure of the dataset is detrimental. This work proposes a novel approach, the Multiple Domain Matching Network (MDMN), to exploit this structure. MDMN embeds all data into a shared feature space while learning which domains share strong statistical relationships. These relationships are often insightful in their own right, and they allow domains to share strength without interference from irrelevant data. This methodology builds on existing distribution-matching approaches by assuming that source domains are varied and outcomes multi-factorial. Therefore, each domain should only match a relevant subset. Theoretical analysis shows that the proposed approach can have a tighter generalization bound than existing multiple-domain adaptation approaches. Empirically, we show that the proposed methodology handles higher numbers of source domains (up to 21 empirically), and provides state-of-the-art performance on image, text, and multi-channel time series classification, including clinical outcome data in an open label trial evaluating a novel treatment for Autism Spectrum Disorder.

Comparative transcriptomics of Entelegyne spiders (Araneae, Entelegynae), with emphasis on molecular evolution of orphan genes.

Next-generation sequencing technology is rapidly transforming the landscape of evolutionary biology, and has become a cost-effective and efficient means of collecting exome information for non-model organisms. Due to their taxonomic diversity, production of interesting venom and silk proteins, and the relative scarcity of existing genomic resources, spiders in particular are excellent targets for next-generation sequencing (NGS) methods. In this study, the transcriptomes of six entelegyne spider species from three genera (Cicurina travisae, C. vibora, Habronattus signatus, H. ustulatus, Nesticus bishopi, and N. cooperi) were sequenced and de novo assembled. Each assembly was assessed for quality and completeness and functionally annotated using gene ontology information. Approximately 100 transcripts with evidence of homology to venom proteins were discovered. After identifying more than 3,000 putatively orthologous genes across all six taxa, we used comparative analyses to identify 24 instances of positively selected genes. In addition, between ~ 550 and 1,100 unique orphan genes were found in each genus. These unique, uncharacterized genes exhibited elevated rates of amino acid substitution, potentially consistent with lineage-specific adaptive evolution. The data generated for this study represent a valuable resource for future phylogenetic and molecular evolutionary research, and our results provide new insight into the forces driving genome evolution in taxa that span the root of entelegyne spider phylogeny.

Multichannel electrophysiological spike sorting via joint dictionary learning and mixture modeling.

We propose a methodology for joint feature learning and clustering of multichannel extracellular electrophysiological data, across multiple recording periods for action potential detection and classification (sorting). Our methodology improves over the previous state of the art principally in four ways. First, via sharing information across channels, we can better distinguish between single-unit spikes and artifacts. Second, our proposed "focused mixture model" (FMM) deals with units appearing, disappearing, or reappearing over multiple recording days, an important consideration for any chronic experiment. Third, by jointly learning features and clusters, we improve performance over previous attempts that proceeded via a two-stage learning process. Fourth, by directly modeling spike rate, we improve the detection of sparsely firing neurons. Moreover, our Bayesian methodology seamlessly handles missing data. We present the state-of-the-art performance without requiring manually tuning hyperparameters, considering both a public dataset with partial ground truth and a new experimental dataset.