Evaluating Metagenomic Quantifications from Next-Generation Sequencing Data

Abstract

Molecular profiling is exploiting the unprecedented power of next generation DNA sequencing to illuminate the microbial diversity of the natural world. The composition of microbiomes has been implicated as an important factor in human health and the function of ecosystems. It is thus of great importance that measurements of microbiomes are accurate and reliable, and moreover it is essential that the accuracy and reliability of such measurements are well understood. This project sought to provide assessments of the accuracy and precision of measurements made by 16S rDNA amplicon sequencing and whole genome shotgun sequencing, as well as investigate the impact of different experimental and bioinformatics choices on quantitative measurements. To address these aims next generation sequencing data from a well quantified metagenomic control material was utilized. Good precision and accuracy were recorded for 16S primer pairs which were perfectly complementary to the target organisms. Where primers were not perfectly complementary to an organism, its abundance was underestimated. Whole genome shotgun sequencing demonstrated very high levels of precision, with a mean coefficient of variation of 2%, and showed good agreement with the 16S rDNA amplicon sequencing using primer pairs optimized specifically for the target species. Small changes in relative species abundance (less than three fold) should be treated with caution as this thesis demonstrated that sequencing results for species can vary by this amount from digital polymerase chain reaction results. Issues with publically available 16S rDNA sequence databases contribute to a lack of taxonomic resolution; taxa measured at low abundance are also likely to be artifacts of the analysis. In addition to the established sequencing platforms, this thesis also investigated the performance of a promising new experimental DNA sequencing platform developed by Oxford Nanopore Technologies (ONT). The ONT MinION, has an error rate of greater than 40% and, while it produces exceptionally long reads, it is not yet suitable for quantitative metagenomics. This thesis also demonstrated that the use of control materials in molecular profiling is important to verify findings and to understand the impact different experimental and bioinformatics choices have on measurements of the microbiome.