Understanding Microbial Community Dynamics in High-Solids Lignocellulolytic Systems Using Bioinformatics Tools

摘要

High-solids lignocellulosic systems are relevant to many different processes including biofuels and global carbon cycling. Due to the high-moisture content of such systems, deconstruction of such lignocellulosic substrates usually occurs due to microbial communities. However, the microbial communities that are responsible for the deconstruction process are poorly understood. Prior work has isolated and characterized individuals within the community, but very little work has sought to characterize the entire community. Using network analysis, an approach only recently applied to microbial communities, this work seeks to characterize and understand the interactions within a high-solids lignocellulosic system.;To understand what a deconstructive community is comprised of and how the structure relates to function, initial work was done to characterize a tomato pomace (seeds and skins of tomato; a common food waste in California) deconstructing community. While the aeration supplied to the microbial communities found no significant differences in the activity of extracted enzymes, the use of a metagenome prediction tool, PICRUSt, suggested differences in types of enzymes present. Hemicellulases tended to enrich among communities that received no aeration while ligninases enriched in communities that were aerated. The control of the environment allowed us to start to identify differences prior to trying to understand an environment like biosolarization where the oxygen concentration is not controlled.;To understand the interaction of the biosolarization microbial community, improvements in network analysis had to be conducted. A new approach to building a microbial network was developed on the principal idea that removing sparsity in favor of smaller sample sizes that would be recombined post-detection was more beneficial to detection of true microbial interactions. If a dataset contained no sparsity, splitting of the dataset performed only slightly worse than the detection method. However, as sparsity in a dataset increases, splitting the dataset into smaller sample sizes to remove sparsity improved the detection of the microbial. This approach could then be used to combine datasets from multiple environments and evaluate microbial interactions on a much larger scale.;Using the new method developed for network analysis, the microbial interactions during biosolarization were studied. Biosolarization is the application of an organic amendment, usually lignocellulosic in nature, to the soil prior to wetting the soil to field capacity and covering with a clear plastic tarp to allow for solar heating. The goal of solarization is to suppress pathogens and inactivate weeds; this is thought to be largely accomplished via the production of volatile fatty acids (VFAs). Work on biosolarization microbial communities has suggested large shifts in the community itself but little was understood as to why the community was shifting and what genes were present during the deconstruction process. Network analysis of biosolarization revealed ten different sub-communities that identify with certain substrates and depths. The sub-communities were tested for correlation to accumulated VFAs and this suggested certain sub-communities may be responsible production of such VFAs. Having identified a sub-community potentially vital to the production of VFAs during biosolarization, investigation into the predictability of such microbial communities was of interest.;The construction of an artificial neural network to predict biosolarization microbial communities was evaluated using lab-based samples. Lab-based sampling show high consistency among a leave-one-out validation process, but prediction of the field samples showed greater error. The error was likely attributed to calculated oxygen concentration during the field biosolarization as a follow up to this work suggested that a change in the assumed oxygen concentration could decrease the error in prediction of the field studies. This suggests that microbial communities do offer a high degree of predictability, but identifying the critical pieces of data that control the microbial community must be accurately targeted and measured. This interplay between the environment and the community leave open many avenues of research such as microbial biomarkers, synthetically constructed communities, and control of communities for things like industrial bioprocessing.