Analysis Methods for Shotgun Metagenomics

The development of whole metagenome shotgun sequencing (WGS) has enabled the precise characterization of taxonomic diversity and functional capabilities of microbial communities in situ while obviating organism isolation and cultivation procedures. WGS created with second- and third-generation sequencing technologies will generate millions of reads and tens (or hundreds) of gigabytes of information about the organisms under investigation. Despite containing an immense amount of information, the reads are unorganized and unlabeled, leading to a significant challenge in discerning from which genome a read originated. Thus, analysis of WGS data necessitates first determining community structure and function from the raw reads before the focus can shift to making multi-sample comparisons. A typical WGS workflow consists of read assignment (taxonomic binning and classification), preprocessing techniques (normalization, dimensionality reduction), exploratory approaches (feature selection and extraction, ordination), statistical inference (regression, constrained ordination, differential abundance analysis), and machine learning. The following chapter provides an overview of these analytical approaches (including challenges and possible pitfalls that may be encountered by researchers) as well as steps toward their solutions. Relevant software packages and resources are also discussed.