PhD Proposal Defense: Adrienne Hoarfrost

The Ph.D. Proposal Defense of Adrienne Hoarfrost will be presented before the UNC Marine Sciences Department in seminar room G201 of Murray Hall and live broadcast to UNC IMS (Institute of Marine Science) seminar room 222 in Morehead City – NC. Held on Monday, April 24th, 2017 from 12:20 pm until 1:20 pm.

Title: Global biogeochemical functional patterns in microbial communities using machine learning approaches with large-scale data integration and computation

Abstract: Microbial communities are an intrinsic component of biogeochemical and ecological function. The functioning of microbial communities, and biogeographical differences in function across space, have global-scale effects on the cycling of carbon and many other elements in the Earth system. However, our ability to describe the functional structure of microbial communities that underpin global-scale processes, and to model the relationship between microbial communities and environmental variables, is still limited. Additionally, a lack of flexible resources for integration of microbial information hinder comparative analyses on a massive or global scale. New approaches to reveal the underlying functional and biogeochemical structure of microbial communities are needed, as well as tools to facilitate curation of and computation on global-scale microbial datasets. I propose to apply a semi-supervised topic modeling approach to identify biogeochemically-relevant clusters of genes from a global dataset of ~3000 soil shotgun metagenomes. These functional clusters will be used to identify functional ecotypes within the global microbial pool, and to investigate the genetic structure underlying biogeochemical patterns. Multivariate models predicting the functional cluster weights from environmental variables will be utilized to predict functional clusters and microbial community metagenomic composition at a global scale, laying the groundwork for incorporation of microbial information into Earth system models. To accomplish these tasks, a global dataset of ~3000 soil shotgun metagenomes will be curated from the major sequencing data repositories, and a non-redundant soil gene reference catalog will be created. Such a gene reference will enable fundamental questions about the diversity of soil microbial communities to be addressed and will provide a crucial resource for soil comparative metagenomics in the future. Finally, a sequencing data discovery tool is under development which enables curation of large-scale sequencing datasets for meta-analyses in an easy-to-use web application, allowing investigators to access data, share curated datasets and improve the existing sequencing metadata pool. The proposed projects together form a multi-pronged approach that will address central questions about microbial diversity, the functional structure driving microbial communities, and their effect on biogeochemical processes, while providing essential tools to facilitate integrative meta-analyses that will accelerate future scientific progress in this expanding and cross-disciplinary field.