Computational Analyses of Metagenomic Data

Abstract

Metagenomics studies the collective microbial genomes extracted from a particular environment without requiring the culturing or isolation of individual genomes, addressing questions revolving around the composition, functionality, and dynamics of microbial communities. The intrinsic complexity of metagenomic data and the diversity of applications call for efficient and accurate computational methods in data handling. In this thesis, I present three primary projects that collectively focus on the computational analysis of metagenomic data, each addressing a distinct topic.

In the first project, I designed and implemented an algorithm named Mapbin for reference-free genomic binning of metagenomic assemblies. Binning aims to group a mixture of genomic fragments based on their genome origin. Mapbin enhances binning results by building a multilayer network that combines the initial binning, assembly graph, and read-pairing information from paired-end sequencing data. The network is further partitioned by the community-detection algorithm, Infomap, to yield a new binning result. Mapbin was tested on multiple simulated and real datasets. The results indicated an overall improvement in the common binning quality metrics.

The second and third projects are both derived from ImMiGeNe, a collaborative and multidisciplinary study investigating the interplay between gut microbiota, host genetics, and immunity in stem-cell transplantation (SCT) patients. In the second project, I conducted microbiome analyses for the metagenomic data. The workflow included the removal of contaminant reads and multiple taxonomic and functional profiling. The results revealed that the SCT recipients' samples yielded significantly fewer reads with heavy contamination of the host DNA, and their microbiomes displayed evident signs of dysbiosis. Finally, I discussed several inherent challenges posed by extremely low levels of target DNA and high levels of contamination in the recipient samples, which cannot be rectified solely through bioinformatics approaches.

The primary goal of the third project is to design a set of primers that can be used to cover bacterial flagellin genes present in the human gut microbiota. Considering the notable diversity of flagellins, I incorporated a method to select representative bacterial flagellin gene sequences, a heuristic approach based on established primer design methods to generate a degenerate primer set, and a selection method to filter genes unlikely to occur in the human gut microbiome. As a result, I successfully curated a reduced yet representative set of primers that would be practical for experimental implementation.