Date of Award

2019

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

First Advisor

Dr. Scott H. Harrison

Abstract

Cost-effectiveness modeling accounts for how expenditures impact outcomes and is an appropriate step towards efficacy of the different methods used for modeling the dynamics of microbial communities. This will help to identify challenging aspects of microbiome studies and the associated costs, including the major differences in research designs (cross-sectional or time series-based) used for conducting such studies. The two major stages of our investigation were to first collect and model cost variable data for microbiome investigations, and then to evaluate how trade-offs related to sample size and expenditures impact investigational outcomes. We screened different potential sources of data for microbiome investigations, ultimately electing to query the NIH grant award database using the NIH RePORTER tool, documenting queries using the PRISMA flow chart, and constraining and defining the final dataset to be used for further statistical analysis. We have implemented a novel integrated framework for assembling a set of different microbiome studies by identifying predictor variables in these studies and effectively collecting and comparing costs across these studies. We next investigated linear and non-linear interactions between predictor variables of cost and constructed a cost model using regression-based models and ensemble models for supervised learning using regression. Finally, we have taken further steps to determine the impact of cost-cutting upon a microbiome study in terms of network recovery and in terms of changes in the microbial interaction networks due to perturbations such as diet. This work overall helps to provide insight on the decisions and outcomes that surround the scaling, planning and implementation of microbiome investigations.

Download

Share

COinS