Please use this identifier to cite or link to this item:
http://hdl.handle.net/11375/27734
Title: | Comprehensive Mapping of Bacterial Metabolism Guides Inquiry Into Specialized Metabolites |
Authors: | Blaga, Victor |
Advisor: | Magarvey, Nathan |
Department: | Biochemistry and Biomedical Sciences |
Publication Date: | 2022 |
Abstract: | Bacterial specialized metabolites (SMs) have long interested scientists due to their diverse chemistries which can harbour antimicrobial activity. The discovery of these molecules experienced a period of exponential success in the mid-1900s and today is guided by next-generation sequencing and liquid chromatograph-mass spectrometry technology. Despite these technological advances, however, they remain under-leveraged. Now, merging disease and antibiotic resistance threaten the security provided by existing antimicrobial medicines. There is an urgent need for a more targeted approach in the discovery of novel SMs, which leverages the tremendous efforts of the past alongside modern big data analytics. To this end, I set out to develop a foundation that could guide SM discovery in the future. I began by bridging all available bacterial metabolism data into a common latent space. Using known metabolic pathways, I generated a library of biosynthetic units used in a novel program to encode metabolites. Each unit was connected to its requisite genes, which were leveraged to reverse engineer this platform by predicting the chemical structures of SMs directly from their encoding genes. Finally, deep learning models were used to annotate some of these chemical-genomic connections as being associated with a particular activity, in this case siderophore activity. This suite of software tools offers promising opportunities to pursue downstream applications, including the connection of unknown genes to metabolites and the identification of novel chemical-genomic connections. In order for these prospects to be realized, however, this intricately-connected network of data necessitates more sophisticated interpretation. Indeed, interconnected chemical, genomic and activity data lends itself particularly well to analysis by graph neural networks. The foundational work described in this communication builds the basis for this comprehensive analysis, which may uncover new insights into the bacterial metabolism we thought we knew. |
URI: | http://hdl.handle.net/11375/27734 |
Appears in Collections: | Open Access Dissertations and Theses |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
blagav_victor_finalsubmission202207_msc.pdf | 85.82 MB | Adobe PDF | View/Open |
Items in MacSphere are protected by copyright, with all rights reserved, unless otherwise indicated.