Inference of In Situ Microbial Physiologies via Sparse Tensor Decomposition of Metatranscriptomes: Application to Cyanobacteria Populations in the North Pacific

Abstract

Microbes respond to changes in their environment by adjusting their physiology through shifts in gene expression, which can be measured in the field by whole community RNA sequencing. The resulting metatranscriptomic data is inherently noisy, with unknown gene functions and fluctuations in organism abundance, all of which limit the utility of traditional methods. In the first chapter of this dissertation, I developed a novel statistical approach that uses sparse tensor decomposition to uncover patterns of gene co-expression. In the second chapter, I applied the method to metatranscriptomic data collected in the North Pacific, focusing on marine cyanobacteria, a group of highly abundant microbes that are responsible for up to a quarter of photosynthesis in global oceans. The analysis uncovered 25 robust co-expression patterns, including four that clarified how cyanobacteria respond in nature to scarce nitrogen and iron nutrients. In the final chapter I looked into another co-expression pattern that revealed how cyanobacteria respond to viral infection, placing this in the context of population diversity and evolution. Altogether this dissertation demonstrates the power of a new analytical approach to elucidate: 1) the functions of unknown genes, 2) how different organisms respond to environmental pressures, and 3) the ways in which microbial physiology and biogeochemical cycles interconnect in a changing ecosystem.