BCCMA: Cognitive phenotypes of emotion regulation and suicide risk in Veterans with TBI

Overall Research Strategy: This submission is part of a BLR&D/CSR&D Collaborative Merit Award for traumatic brain injury (TBI; BCCMA) proposal (RFP #BX-24-007) involving four separate but integrated submissions that together investigate the overarching model that TBI impacts cognitive control and emotion dysregulation via multiple pathways, which in turn increase risk of suicide. Since this pathway cannot be fully evaluated by a single study, it will be translationally evaluated across sites with shared assessments in humans and across species. This Project: Within the broader BCCMA, this specific project will leverage existing data collected in the prior phase of this collaboration, which was carefully designed to collect parallel measures in animals (before and after TBI) and humans (with and without TBI and/or history of suicidality), including behavioral tests of attention and impulsivity, blood markers of inflammation, and measures of brain integrity and dysfunction. We will apply a suite of computational modeling methods to both animal and human data. First, starting with behavioral data collected from rats and Veterans, we will use computational modeling to “customize” or fit models of the decision-making process to closely replicate the behavior of each individual rat or human on laboratory tasks (Aim 1). The model generates latent cognitive variables not directly evident from the behavior alone. This will allow us to examine whether these latent cognitive variables change in the same way after TBI in rats as in humans – either strengthening the animal model or else pointing out places where it needs to be modified to more closely parallel humans. Second, using the full set of behavioral, biological, and modeling variables available from our collaborative project, we will develop machine learning (ML) models to distinguish TBI vs. non-TBI in both rats and humans (Aim 2). This will allow us to determine cognitive and biological processes that are most reliably disrupted by TBI in the rat model and in humans; if there is high overlap, then this again strengthens the animal model and points to specific variables that should be targeted for therapeutic intervention in pre-clinical trials; if there is poor overlap, then this will signal variables in the animals that need to be prioritized for study in humans, or vice versa. We will also use ML methods to predict which variables, including long-term effects of TBI such as inflammation, are most associated with suicidality in humans (Aim 3). The results of this project will provide much-needed information on common phenotypes observed in Veterans with TBI and in animal models of TBI, facilitating translational application. The predictive models will help identify symptoms and markers, common to both animals and humans, that can signal long-term changes in the wake of TBI, including inflammation and cognitive changes that could increase risk for suicide. A key focus of this work is that there may not be a single set of markers that describes all at-risk individuals, but rather multiple pathways to suicide risk, involving different configurations of symptoms. If successful, the predictive model could be deployed to aid identification of Veterans for whom TBI has led to a series of long-term changes increasing risk for suicide, so that clinical resources can be appropriately targeted to help those at highest risk. Additionally, the project’s emphasis on common cognitive and physiological endophenotypes in animals and humans will aid translational application, as the same targets can be modified in pre-clinical trials as would ultimately be targeted in clinical trials.