Deep Phenotyping of Behavioral Patterns Associated with Neuropsychiatric Disease
Need/Problem: Why have advances in our understanding and treatment of mental illness lagged behind that of other biomedical conditions? To a large extent, this difference stems from the fundamental issue of diagnosis. For mental illness, diagnosis still relies exclusively on checklists of signs and symptoms, derived from observation by the clinician and self-reports from the patient. While this is critical, we can dramatically improve mental health diagnosis and treatment by fully integrating genetic, physiological, and neurocircuit information. Much of these data rely heavily on preclinical experiments where studying genetic or neurocircuit underpinnings of mental illness. However, in preclinical models there is a clear lack of behavioral readouts that can be 1) associated with a particular psychiatric condition, and 2) clearly linked to back to molecular or circuit mechanisms.
Grant Summary: Rodent behavioral readouts such as locomotion, grooming, and position within the environment are commonly used to infer neuropsychiatric disease symptoms observed in humans. These measurements are typically quantified in low dimensional space (time spent in a location or performing a single behavior), which limits the capacity by which we can assess how neurocircuit perturbations or therapeutics alter global brain and behavior relationships. To address this, machine learning and computer vision approaches that extract meaningful patterns of information from high dimensional video data, produce novel, highly quantitative behavioral metrics. We are currently adapting this technology and plan to generate a detailed database exploring how optogenetic manipulation of critical neuronal circuits produces previously unidentified behavioral patterns. In addition, we will generate a data analysis pipeline which can be used by other researchers at UNC to store and analyze detailed behavioral data collected in any species, including the potential for investigating human behavioral patterns.
Goals and Projected Outcomes: Collectively, we aim to demonstrate the feasibility of this approach in order to submit a large-scale NIH BRAIN Initiative, NIMH, and/or NICHD proposal within 1-2 years. At that time, we hope to have generated a standardized system by which behavioral data is automatically stored, processed, and analyzed. We will also generate a database indexing system in order to make our infrastructure highly scalable and searchable based on a number of experimental parameters. These experiments will greatly improve our understanding of how neurocircuit alterations change behavior and will provide a means by which detailed behavioral metrics associated with specific illnesses can be characterized in preclinical models.
We anticipate that this experiment will produce the following scientific deliverables:
- A dataset that will provide an important quantitative foundation in order to determine how manipulation of precise neurocircuits alters global patterns of behavior.
- Strong preliminary data outlining the feasibility and potential impact of this project. We will leverage this in order to submit a multi-PI large scale grant aimed at generating infrastructure and data analysis capacity that can be applied to study behavioral changes in many species.
Grant Details: We acquired the original version of the behavioral analysis code in August, 2016. We setup the required depth camera for behavioral monitoring in September, 2016. We are currently working to extract behavioral “syllables” from our own collected data as of October 2016. We then plan to conduct behavioral experiments to detail how known psychoactive compounds (acute cocaine, fluoxetine, lorazepam, etc.) alter the distribution and sequence of behaviors. This will serve as an important starting point since these manipulations have already been previously shown to produce gross changes in behavior (reward, locomotion, reduced anxiety, etc.). In parallel, we will perform two experiments where optogenetic manipulations of particular circuits are known. We anticipate having the data acquisition and storage solution for this project completely up and running by January 2017.
Garret D. Stuber, Ph.D.
Need/Problem: Binge-eating disorder (BED) and bulimia nervosa (BN) are serious and costly eating disorders. Our genetic studies on anorexia nervosa are shedding novel light into the metabolic aspects of this psychiatric disorder, but we know essentially nothing about the molecular genetic basis of BN and BED beyond the established twin-based observation that they are highly heritable.
Grant Summary: We will conduct an extremely rapid and large genome-wide association study (GWAS) in 2500 individuals with binge-type eating disorders and matched controls, coupled with five-site microbiota sampling, and deep phenotyping using an eating disorders recovery app programmed on Apple Watches.
Goals & Projected Outcomes: Our goal is to conduct the first and largest study of individuals with binge-type eating disorders that combines genomic and microbiota data with deep phenotyping. Information from these three domains will allow us not only to explore predictors of outcome, but also to build algorithms that predict behavioral events (e.g., impending binges or purges) to be able to intervene in real time via wearable technology.
Grant Details: This project represents a unique academic-industry-foundation partnership including services and materials from Apple (Apple Watches), the National Institute of Mental Health (saliva DNA collection kits), uBiome (5-site microbiota sampling kits and analyses), and partnership with Recovery Record (the largest eating disorders recovery app in the world), and support from the National Eating Disorders Association, the Binge Eating Disorder Association, and the Foundation of Hope.
We are the first study of any illness to combine genome-wide association study with an investigation of microbes across 5 bodily sites, coupled with extensive phenotyping over the period of a month using Apple Watch technology. We will use advanced analytic techniques to combine information from these three domains: genomic, microbiota, and phenotyping to create a rich biological and behavioral picture of binge-type eating disorders. We will then apply that knowledge both to developing novel biologically informed interventions as well as behavioral interventions that utilize the predictive algorithms we develop using the passive and active data collection from the Apple Watches.
BEGIN will allow us to identify genetic and microbiota contributors to disorder risk and maintenance; identify genomic, microbiota, and behavioral predictors of outcome; and to build algorithms that predict behavioral events (e.g., impending binges or purges) to enable real-time intervention via wearable technology.