Dissertation
Novel machine learning frameworks for disease-associated microbiome interaction identification and analysis
University of Iowa
Doctor of Philosophy (PhD), University of Iowa
Spring 2022
DOI: 10.17077/etd.006365
Abstract
Understanding the causes for dysbiosis of oral microbiome is critical to develop diagnostic tools and therapies for a wide range of human oral diseases. The overall goal of our study was to improve microbiome omics data analysis by developing new analytic tools that take the advantages of machine learning based approaches to gain new knowledge from oral microbiome. We designed and implemented three novel computational tools that mine hidden information from oral microbiome omics datasets (i.e., metagenomics and metatranscriptomics) and investigated their association with oral diseases including dental caries and periodontitis.
Three research aims were achieved in this project. The first aim was to develop a novel computational framework to investigate the landscape of microbiome community-wide transcription factor binding motif (CW-TFBM). For the second aim, we developed a machine learning (ML) based computational framework to identify microbiome biomarkers. The third aim was to develop association rule mining (ARM) framework to infer complex and heterogenous interactions in bacteria communities using microbiome multiple meta-omics data. We applied these tools on a microbiome dataset associated with dental caries and periodontitis. Our study revealed, for the first time, the CW-TFBM landscape of oral disease associated microbiome. The biomarker discovery framework identified multiple different feature sets that have potentials for disease prediction and diagnosis. The ARM based interaction discovery framework enabled us to integrate multiple microbiome heterogeneous meta-omics data to gain new knowledge.
In summary, the computational tools resulted from this research revealed unknown aspects of oral microbiomes and help us gain a wholistic understanding of the associations between oral microbiome and oral diseases, thus the outcomes from this project will improve the health and well-being of the public.
Details
- Title: Subtitle
- Novel machine learning frameworks for disease-associated microbiome interaction identification and analysis
- Creators
- Miyuraj Harishchandra Hikkaduwa Withanage
- Contributors
- Erliang Zeng (Advisor)Xian Jin Xie (Committee Member)Huojun Cao (Committee Member)Jeffrey A Banas (Committee Member)David R Drake (Committee Member)
- Resource Type
- Dissertation
- Degree Awarded
- Doctor of Philosophy (PhD), University of Iowa
- Degree in
- Informatics (Bioinformatics and Computational Biology)
- Date degree season
- Spring 2022
- DOI
- 10.17077/etd.006365
- Publisher
- University of Iowa
- Number of pages
- xviii. 215 pages
- Copyright
- Copyright 2022 Miyuraj Harishchandra Hikkaduwa Withanage
- Language
- English
- Description illustrations
- illustrations (chiefly color), tables, graphs
- Description bibliographic
- Includes bibliographical references.
- Public Abstract (ETD)
- A healthy oral microbiome can become dysbiosis and be associated with an oral disease in response to a stimulus (e.g., change of dietary habits, smoking, and onset of certain diseases). An oral microbiome associated with a disease always contains increased abundance of pathogenic microorganisms. This transformation is a community shift. Understanding the causes for the community shift of oral microbiome is critical to develop diagnostic tools and therapies for a wide range of human oral diseases. Methods for understanding this community shift in terms of changes of species composition and gene expression have been established. However, our knowledge on the causes such as regulatory mechanisms governing the gene expression driving this transformation is very limited. To fill this knowledge gap, we have successfully created a series of computational methods to better understand oral microbiomes and their associations with oral diseases. Further, we have developed computational methods to identify discriminatory biomarkers from oral microbiomes that can be used to differentiate patients who have oral diseases from health population. The biomarkers can be gene regulatory elements, species, and genes. It is the change of interactions among many biological players that cause disease. Thus, we have developed a computational framework to infer networks formed by different biomarkers within a microbiome and to compare these networks across microbiomes, which helps us gain a holistic understanding of the roles of microbiome biomarkers to oral diseases. The outcome of this project has the potential to improve the health and well-being of the public.
- Academic Unit
- Craniofacial Anomalies Research Center; IDGP in Informatics
- Record Identifier
- 9984271154002771
Metrics
2 File views/ downloads
122 Record Views