Date of Award

5-19-2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Bioinformatics

First Advisor

David Ussery

Abstract

Tuberculosis (TB) remains a public health threat in Arkansas. Great strides have been made over the past fifty years to reduce the number of TB cases in Arkansas. However, the rate of TB cases has remained steady since 2010. According to the Arkansas Department of Health, from 2012 through 2023 the TB Case rate per 100,00 population was between 2.0 and 3.1. To continue Arkansas’ drive towards elimination of TB in Arkansas, defined as one TB case per million residents, new data analysis tools are needed. This project employs multiple bioinformatic tools to analyze TB cases and latent tuberculosis infections (LTBI) data to gain new insights and deliverable actionable data to the Arkansas Department of Health. The project utilizes Social Network Analysis to visualize and statistically determine the most important TB cases in a large-scale TB outbreak in Southwest Arkansas. Multiple machine learning algorithms, t-SNE, MCMC, and xgboost, are deployed to analyze genotypically unique Arkansas TB isolates, infer whom-infected-whom transmission in the Southwest Arkansas TB outbreak, and identify the most relevant variables contributing to Missed Opportunity Arkansas TB Cases. The project uses spatial statistical analysis to examine if LTBI testing was conducted near TB case “hotspots”. The Social Network Analysis identified three TB Case nodes with high statistical influence in the TB outbreak network, leading to greater scrutiny of the physical places were these nodes visited in Southwest Arkansas. Utilizing SHAP values drawn from the t-SNE algorithm, two distinct cohorts of unique TB genomes were identified. The inference of whom-infected-who identified cases were on-going transmission maybe occurring, as well as predicting missing TB cases, while providing insight on whom the missing TB cases associated with. Using SHAP values derived from the xgboost algorithm, the most important variables associated with missed opportunities were identified, allowing the TB program to focus interventions on persons with these variables. To analyze LTBI testing in spatial space with TB cases, the nation’s first LTBI database was constructed from medical records. Using both spatial statistics and visualizations, it was confirmed that LTBI testing occurs near TB hotspots. These separate projects, joined around a common theme of providing new insight and actionable data, show how many facets of analysis, deploying many bioinformatic tools, can provide new insights into current situations, help guide public health interventions to be more effective and efficient, and help public health decision makers be more strategic in their decision making, all of which may lead to a continued push to reach the TB elimination threshold in Arkansas.

Download

Share

COinS