AI/ML Laboratory for Mental Health and Addiction

The Developmental Neuropsychopharmacology Laboratory investigates mechanisms underlying addiction as well as brain reward pathways associated with nicotine, alcohol, and opioids. The laboratory also focuses on how the impact of early life stress and toxic stress exposure influences later development of substance use disorder (SUD) involving alcohol, nicotine, and opioids. The effects of human immunodeficiency virus (HIV) on the brain parallel those of SUD, with which it is often comorbid. Co-morbidities such as SUD, HIV, and other chronic disorders exacerbate the impact on mental and behavioral health. Thus, the laboratory also investigates the bases of HIV-related neurological conditions (e.g., HIV-associated neurological disorder or HAND). The laboratory has an established record of using rodent models of impaired reward processing to test behavior, pharmacology, genetics, and biochemistry associated with affective disorders and addiction. Recently, the laboratory has acquired medical histories and functional magnetic resonance imaging (fMRI) data from people living with HIV and SUD. We aim to apply AI/ML analysis techniques to social determinants of health and genomics data from these human subjects to predict the likelihood of a given person having SUD and/or HIV. The ultimate long-term goal of our research is to promote brain health through the application of advanced analysis strategies (e.g., AI/ML) and thereby help clinicians assess health disparities that result from social determinants of health and ancestry-specific genetic variations not considered in most treatment strategies.

We have acquired resting state functional magnetic resonance imaging (fMRI) and diffusion tensor imaging (DTI) of people living with HIV but not SUD, people who have SUD but not HIV, and comorbid populations. Data analyses will emphasize limbic structures (e.g., amygdala, hippocampus, nucleus accumbens, etc.) given their roles in addiction and socialization.

We have acquired genomic and social determinants of health data from people living with HIV but not SUD, people who have SUD but not HIV, and comorbid populations. Data analyses will employ LASSO-CV and Consensus Feature Elimination machine learning algorithms to predict who is more likely to have HIV, SUD, or both.

We aim to enhance the predictive power of our AI/ML-based analyses by integrating neuroimaging data extracted from people living with HIV, people who have SUD, and comorbid populations with their corresponding genomic and social determinants of health data. AI/ML algorithms will be selected based on their ability to optimize predictive power.