Examining NeuroHIV-Induced Immune Responses Through Gene Expression and a Newly Designed Machine Learning-Based Tool for Automatic Image Analysis
Abstract
Human Immunodeficiency Virus (HIV) remains a global health concern even as antiretroviral therapies (ART) improve livelihood outcomes for infected individuals. Viral infection in the brain presents a reservoir removed from the effects of these treatments. As the population of HIV infected individuals ages with greater longevity, neurocognitive decline has remained prevalent despite ART. Addressing this neurocognitive decline requires altering the chronic immuno-activation that appears to drive the dysfunction and developing interventions within a complete understanding of immune pathways involved.
This dissertation seeks to provide novel insight into genes involved in HIV induced neurodegeneration as well as introducing a program designed to enhance the throughput of this and similar studies on dynamic cellular systems. ACCT: Automatic Cell Counting with Trainable Weka Segmentation was developed to be accessible in both computational and computer science expertise requirements and we have compared its performance to similar programs on neurological datasets. While we developed this program using data from neuroscience, it stands as a broadly accessible and useful tool across any imaging data with sufficiently defined objects to be quantified. The primary resident immune cells of the brain are the microglia. These cells champion the innate immune response to viruses via the type 1 and 2 interferon response. Interferon response genes (IRG), however, also have the capability of feeding back on the pathway itself to modify the throughput of the immune response. Two genes which have regulatory effects on type 1 interferon response are Interferon Regulatory Factor 7 (IRF7), a promoter of induction and enhancement of type 1 interferon activity and Guanylate Binding Protein 4 (GBP4), an inhibitor of IRF7 function.
I have associated the ortholog to this GBP gene with immune response to HIV in an HIV infection model using gene expression data with population specific enrichment analysis (PSEA). I have also examined this protein’s localization in mouse cortex, and observed distinct staining populations of immune cells. This dissertation seeks to advance understanding through increased understanding of HIV-induced neurodegeneration. The open-access design of ACCT is intended to increase researchers’ ability to perform similar analyses using this computational tool that we have designed and contextualized within and against similar automatic image quantification methods.