Simple Behavioral Analysis (SimBA) as a platform for explainable machine learning in behavioral neuroscience.

The study of complex behaviors is often challenging when using manual annotation due to the absence of quantifiable behavioral definitions and the subjective nature of behavioral annotation. Integration of supervised machine learning approaches mitigates some of these issues through the inclusion of accessible and explainable model interpretation. To decrease barriers to access, and with an emphasis on accessible model explainability, we developed the open-source Simple Behavioral Analysis (SimBA) platform for behavioral neuroscientists. SimBA introduces several machine learning interpretability tools, including SHapley Additive exPlanation (SHAP) scores, that aid in creating explainable and transparent behavioral classifiers. Here we show how the addition of explainability metrics allows for quantifiable comparisons of aggressive social behavior across research groups and species, reconceptualizing behavior as a sharable reagent and providing an open-source framework. We provide an open-source, graphical user interface (GUI)-driven, well-documented package to facilitate the movement toward improved automation and sharing of behavioral classification tools across laboratories.

Individual Differences in Volitional Social Self-Administration and Motivation in Male and Female Mice Following Social Stress.

BACKGROUND: A key challenge in developing treatments for neuropsychiatric illness is the disconnect between preclinical models and the complexity of human social behavior. We integrate voluntary social self-administration into a rodent model of social stress as a platform for the identification of fundamental brain and behavior mechanisms underlying stress-induced individual differences in social motivation. METHODS: Here, we introduced an operant social stress procedure in male and female mice composed of 3 phases: 1) social self-administration training, 2) social stress exposure concurrent with reinforced self-administration testing, and 3) poststress operant testing under nonreinforced and reinforced conditions. We used social-defeat and witness-defeat stress in male and female mice. RESULTS: Social defeat attenuated social reward seeking in males but not females, whereas witness defeat had no effect in males but promoted seeking behavior in females. We resolved social stress-induced changes to social motivation by aggregating z-scored operant metrics into a cumulative social index score to describe the spectrum of individual differences exhibited during operant social stress. Clustering does not adequately describe the relative distributions of social motivation following stress and is better described as a nonbinary behavioral distribution defined by the social index score, capturing a dynamic range of stress-related alterations in social motivation inclusive of sex as a biological variable. CONCLUSIONS: We demonstrated that operant social stress can detect stable individual differences in stress-induced changes to social motivation. The inclusion of volitional behavior in social procedures may enhance the understanding of behavioral adaptations that promote stress resiliency and their mechanisms under more naturalistic conditions.

An arginine-rich nuclear localization signal (ArgiNLS) strategy for streamlined image segmentation of single-cells.

High-throughput volumetric fluorescent microscopy pipelines can spatially integrate whole-brain structure and function at the foundational level of single-cells. However, conventional fluorescent protein (FP) modifications used to discriminate single-cells possess limited efficacy or are detrimental to cellular health. Here, we introduce a synthetic and non-deleterious nuclear localization signal (NLS) tag strategy, called 'Arginine-rich NLS' (ArgiNLS), that optimizes genetic labeling and downstream image segmentation of single-cells by restricting FP localization near-exclusively in the nucleus through a poly-arginine mechanism. A single N-terminal ArgiNLS tag provides modular nuclear restriction consistently across spectrally separate FP variants. ArgiNLS performance in vivo displays functional conservation across major cortical cell classes, and in response to both local and systemic brain wide AAV administration. Crucially, the high signal-to-noise ratio afforded by ArgiNLS enhances ML-automated segmentation of single-cells due to rapid classifier training and enrichment of labeled cell detection within 2D brain sections or 3D volumetric whole-brain image datasets, derived from both staining-amplified and native signal. This genetic strategy provides a simple and flexible basis for precise image segmentation of genetically labeled single-cells at scale and paired with behavioral procedures.

Temporal scaling of dopamine neuron firing and dopamine release by distinct ion channels shape behavior.

Dopamine is broadly implicated in reinforcement learning, but how patterns of dopamine activity are generated is poorly resolved. Here, we demonstrate that two ion channels, Kv4.3 and BKCa1.1, regulate the pattern of dopamine neuron firing and dopamine release on different time scales to influence separate phases of reinforced behavior in mice. Inactivation of Kv4.3 in VTA dopamine neurons increases ex vivo pacemaker activity and excitability that is associated with increased in vivo firing rate and ramping dynamics before lever press in a learned instrumental paradigm. Loss of Kv4.3 enhances performance of the learned response and facilitates extinction. In contrast, loss of BKCa1.1 increases burst firing and phasic dopamine release that enhances learning of an instrumental response and enhances extinction burst lever pressing in early extinction that is associated with a greater change in activity between reinforced and unreinforced actions. These data demonstrate that disruption of intrinsic regulators of neuronal activity differentially affects dopamine dynamics during reinforcement and extinction learning.

Oxytocin receptor is not required for social attachment in prairie voles.

Prairie voles are among a small group of mammals that display long-term social attachment between mating partners. Many pharmacological studies show that signaling via the oxytocin receptor (Oxtr) is critical for the display of social monogamy in these animals. We used CRISPR mutagenesis to generate three different Oxtr-null mutant prairie vole lines. Oxtr mutants displayed social attachment such that males and females showed a behavioral preference for their mating partners over a stranger of the opposite sex, even when assayed using different experimental setups. Mothers lacking Oxtr delivered viable pups, and parents displayed care for their young and raised them to the weanling stage. Together, our studies unexpectedly reveal that social attachment, parturition, and parental behavior can occur in the absence of Oxtr signaling in prairie voles.

Sex differences in appetitive and reactive aggression.

Aggression is an evolutionarily conserved, adaptive component of social behavior. Studies in male mice illustrate that aggression is influenced by numerous factors including the degree to which an individual finds aggression rewarding and will work for access to attack and subordinate mice. While such studies have expanded our understanding of the molecular and circuit mechanisms of male aggression very little is known about female aggression, within these established contexts. Here we use an ethologically relevant model of male vs. female aggression by pair housing adult male and female outbred CFW mice with opposite sex cage mates. We assess reactive (defensive) aggression in the resident intruder (RI) test and appetitive (rewarding) aggression in the aggression conditioned place preference (CPP) and operant self-administration (SA) tests. Our results show dramatic sex differences in both qualitative and quantitative aspects of reactive vs. appetitive aggression. Males exhibit more wrestling and less investigative behavior during RI, find aggression rewarding, and will work for access to a subordinate to attack. Females exhibit more bites, alternate between aggressive behaviors and investigative behaviors more readily during RI, however, they do not find aggression to be rewarding or reinforcing. These results establish sex differences in aggression in mice, providing an important resource for the field to better understand the circuit and molecular mechanisms of aggression in both sexes.

Toward the explainability, transparency, and universality of machine learning for behavioral classification in neuroscience.

The use of rigorous ethological observation via machine learning techniques to understand brain function (computational neuroethology) is a rapidly growing approach that is poised to significantly change how behavioral neuroscience is commonly performed. With the development of open-source platforms for automated tracking and behavioral recognition, these approaches are now accessible to a wide array of neuroscientists despite variations in budget and computational experience. Importantly, this adoption has moved the field toward a common understanding of behavior and brain function through the removal of manual bias and the identification of previously unknown behavioral repertoires. Although less apparent, another consequence of this movement is the introduction of analytical tools that increase the explainabilty, transparency, and universality of the machine-based behavioral classifications both within and between research groups. Here, we focus on three main applications of such machine model explainabilty tools and metrics in the drive toward behavioral (i) standardization, (ii) specialization, and (iii) explainability. We provide a perspective on the use of explainability tools in computational neuroethology, and detail why this is a necessary next step in the expansion of the field. Specifically, as a possible solution in behavioral neuroscience, we propose the use of Shapley values via Shapley Additive Explanations (SHAP) as a diagnostic resource toward explainability of human annotation, as well as supervised and unsupervised behavioral machine learning analysis.

Rage Against the Machine: Advancing the study of aggression ethology via machine learning.

RATIONALE: Aggression, comorbid with neuropsychiatric disorders, exhibits with diverse clinical presentations and places a significant burden on patients, caregivers, and society. This diversity is observed because aggression is a complex behavior that can be ethologically demarcated as either appetitive (rewarding) or reactive (defensive), each with its own behavioral characteristics, functionality, and neural basis that may transition from adaptive to maladaptive depending on genetic and environmental factors. There has been a recent surge in the development of preclinical animal models for studying appetitive aggression-related behaviors and identifying the neural mechanisms guiding their progression and expression. However, adoption of these procedures is often impeded by the arduous task of manually scoring complex social interactions. Manual observations are generally susceptible to observer drift, long analysis times, and poor inter-rater reliability, and are further incompatible with the sampling frequencies required of modern neuroscience methods. OBJECTIVES: In this review, we discuss recent advances in the preclinical study of appetitive aggression in mice, paired with our perspective on the potential for machine learning techniques in producing automated, robust scoring of aggressive social behavior. We discuss critical considerations for implementing valid computer classifications within behavioral pharmacological studies. KEY RESULTS: Open-source automated classification platforms can match or exceed the performance of human observers while removing the confounds of observer drift, bias, and inter-rater reliability. Furthermore, unsupervised approaches can identify previously uncharacterized aggression-related behavioral repertoires in model species. DISCUSSION AND CONCLUSIONS: Advances in open-source computational approaches hold promise for overcoming current manual annotation caveats while also introducing and generalizing computational neuroethology to the greater behavioral neuroscience community. We propose that currently available open-source approaches are sufficient for overcoming the main limitations preventing wide adoption of machine learning within the context of preclinical aggression behavioral research.

Affiliation, Aggression, and Selectivity of Peer Relationships in Meadow and Prairie Voles.

Relationships between adult peers are central to the structure of social groups. In some species, selective preferences for specific peers provide a foundation for consistent group composition. These preferences may be shaped by affiliation toward familiar individuals, and/or by aversion to unfamiliar individuals. We compared peer interactions in two vole species that form selective preferences for familiar same-sex individuals but differ in mating system. Prairie voles (Microtus ochrogaster) form pair bonds with mates and may reside in family groups. Meadow voles (Microtus pennsylvanicus) are promiscuous breeders that form communal winter groups in the wild, and exhibit greater social behavior in short day (SD) lengths in the laboratory. We characterized affiliative, anxiety-like, and aggressive interactions with familiar and novel same-sex conspecifics in meadow and prairie voles housed in summer- or winter-like photoperiods. Species differences in affective behaviors were pronounced, with prairie voles exhibiting more aggressive behavior and less anxiety-like behavior relative to meadow voles. Meadow voles housed in short (vs. long) day lengths were more affiliative and more interactive with strangers; prosocial behavior was also facilitated by a history of social housing. Prairie voles exhibited partner preferences regardless of sex or day length, indicating that selective peer preferences are the norm in prairie voles. Prairie vole females formed preferences for new same-sex social partners following re-pairing; males were often aggressive upon re-pairing. These data suggest that preferences for familiar peers in prairie voles are maintained in part by aggression toward unfamiliar individuals, as in mate partnerships. In contrast, social tolerance is an important feature of meadow vole peer affiliation, demonstrated by low aggression toward unfamiliar conspecifics, and consistent with field data on winter tolerance.

Comparative role of reward in long-term peer and mate relationships in voles.

This is a contribution to SI: SBN/ICN meeting. In social species, relationships may form between mates, parents and their offspring, and/or social peers. Prairie voles and meadow voles both form selective relationships for familiar same-sex peers, but differ in mating system, allowing comparison of the properties of peer and mate relationships. Prairie vole mate bonds are dopamine-dependent, unlike meadow vole peer relationships, indicating potential differences in the mechanisms and motivation supporting these relationships within and/or across species. We review the role of dopamine signaling in affiliative behavior, and assess the role of behavioral reward across relationship types. We compared the reinforcing properties of mate versus peer relationships within a species (prairie voles), and peer relationships across species (meadow and prairie voles). Social reinforcement was assessed using the socially conditioned place preference test. Animals were conditioned using randomly assigned, equally preferred beddings associated with social (CS+) and solitary (CS-) housing. Prairie vole mates, but not prairie or meadow vole peers, conditioned toward the social cue. A second study in peers used counter-conditioning to enhance the capacity to detect low-level conditioning. Time spent on CS+ bedding significantly decreased in meadow voles, and showed a non-significant increase in prairie voles. These data support the conclusion that mate relationships are rewarding for prairie voles. Despite selectivity of preferences for familiar individuals in partner preference tests, peer relationships in both species appear only weakly reinforcing or non-reinforcing. This suggests important differences in the pathways underlying these relationship types, even within species.

Stress impairs new but not established relationships in seasonally social voles.

Affiliative social relationships are impacted by stressors and can shape responses to stress. However, the effects of stress on social relationships in different contexts are not well understood. Meadow voles provide an opportunity to study these effects on peer relationships outside of a reproductive context. In winter months, female meadow voles cohabit with peers of both sexes, and social huddling is facilitated by exposure to short, winter-like day lengths in the lab. We investigated the role of stress and corticosterone (cort) levels in social behavior in short day-housed female meadow voles. A brief forced swim elevated cort levels, and we assessed the effects of this stressor on new and established relationships between females. In pairs formed following exposure to swim stress, the stressor significantly reduced the fraction of huddling time subjects spent with a familiar partner. Swim stress did not affect partner preferences in pairs established prior to the stressor. Finally, we examined fecal glucocorticoid metabolite levels via immunoassay in voles housed under short day (10h light) versus long day (14 h light) conditions and detected higher glucocorticoid levels in long day-housed voles. These findings support a role for stress regulation in the formation of social relationships in female meadow voles, and are consistent with a potential role for seasonal variation in cort in the behavioral transition from solitary to social. Together they highlight the importance of stress and possibly glucocorticoid signaling for social behavior.

Sex, social status, and CRF receptor densities in naked mole-rats.

Naked mole-rats (Heterocephalus glaber) live in groups that are notable for their large size and caste structure, with breeding monopolized by a single female and a small number of males. Recent studies have demonstrated substantial differences between the brains of breeders and subordinates induced by changes in social standing. Corticotropin-releasing factor (CRF) receptors-which bind the hormone CRF as well as related peptides-are important regulators of stress and anxiety, and are emerging as factors affecting social behavior. We conducted autoradiographic analyses of CRF1 and CRF2 receptor binding densities in female and male naked mole-rats varying in breeding status. Both globally and in specific brain regions, CRF1 receptor densities varied with breeding status. CRF1 receptor densities were higher in subordinates across brain regions, and particularly in the piriform cortex and cortical amygdala. Sex differences were present in CRF2 receptor binding densities, as is the case in multiple vole species. CRF2 receptor densities were higher in females, both globally and in the cortical amygdala and lateral amygdalar nucleus. These results provide novel insights into the neurobiology of social hierarchy in naked mole-rats, and add to a growing body of work that links changes in the CRF system with social behavior.