Haozheng Tian, Bioinformatics Thesis Defense

In partial fulfillment of the requirements for the degree of Doctor of Philosophy in Bioinformatics in the School of Biological Sciences Haozheng Tian Defended his thesis: Public Health Informatics - Strategy and Decision Modeling Thursday, August 8th, 2019 1:30 PM Eastern Time Klaus 1116 East Thesis Advisor: Dr. Eva K. Lee H. Milton Stewart School of Industrial and Systems Engineering Georgia Institute of Technology Committee Members: Dr. King Jordan School of Biological Sciences Georgia Institute of Technology Dr. John F. McDonald School of Biological Sciences Georgia Institute of Technology Dr. Jung H. Choi School of Biological Sciences Georgia Institute of Technology Dr. Concettina Guerra College of Computing Georgia Institute of Technology Dr. Peng Qiu Department of Biomedical Engineering Georgia Institute of Technology Abstract: My research is composed of three studies focused on providing decision modeling and analytical tools with the objective of protecting public health. The first study introduces an agent-based simulation platform that serves as a decision support system for crowd management in public venues. I propose a new implementation of agent-based simulation with improvement on four aspects: path planning, collision avoidance, emotion modeling and optimization with simulation. The deliverables of this study also include a complete simulation platform for researcher's use. The second study applies a data-driven informatics and machine learning approach to quantify the outcome of practice variance of medical care providers. The study investigates the safety and efficacy of a large-dose, needle-based epidural anesthesia technique for parturient women. Machine learning model is proposed as the classifier to predict the occurrence of hypotension. Further, machine learning approach is applied to predict the outcome of epidural anesthesia, uncovering the important factors of a successful practice. Quantification of the effect of practice variance and medicine usage is provided. The findings from this investigation facilitate delivery improvement and establish an improved clinical practice guideline for training and for dissemination of safe practice. The third study proposes the application of convolutional neural network (CNN) in the prediction of antigenicity of influenza viruses (A/H3N2) and vaccine recommendation. The study systematically explores the ways of representation of hemagglutinin (HA) besides using binary digit or character as widely applied in other researches. Heuristic optimization is applied to optimize the selection of AAindex as well as the structure of CNN. Contrasting to other state-of-the-art approaches, the model offers better coverage in vaccine recommendation and have superior performance in accurate prediction of antigenicity.