Metagenomic Approaches for Determining Low-Abundance Genomic Entities in Microbiomes

Metagenomics is one of the most promising techniques of the last decade to make pathogen detection faster and test for all pathogen in a single analysis. However, discovery and accurate identification of low-abundance entities in metagenomic samples remain challenging; especially since each sample type possess its own properties that need to be taken into account during the study design, for example a high abundance of host sequences or the quality of reference sequences that represent the genomes of organisms.

In the first study we investigated the prospect of utilizing metagenomics for pathogen detection in patients with endophthalmitis, an intraocular eye infection. Vitreous, the intraocular fluid, was collected from 14 patients with endophthalmitis and sequenced. Several types of negative controls were also analyzed: a) samples from patients without eye infection, b) balanced salt solutions before and after loading onto the surgical instrument, and c) DNA extraction blank controls (vials without added sample).

We examined the influence of contaminations on the results and how to mitigate them. For this we utilized two different DNA isolation procedures, since every DNA extraction kit has its unique fingerprint of contaminant DNA present in the reagents. We found that using an ultra-clean kit was better suited to work with vitreous fluid.

Additionally, we detected classification errors caused by contaminant sequences in public reference genomes and attempted to address these errors by removing contaminant DNA sequences in reference genomes as well as read sequences of human origin prior to the analysis. Through careful analysis we were able to confirm 11 out of 12 culture positive cases with our metagenomic approach.

If difficulties relating to sample and reference genome contamination can be solved, it should be possible to differentiate between sterile and infectious endophthalmitis with our presented protocol. The goal of the second study was to evaluate the feasibility of detecting parasites in fecal samples animals (i.e. pigs and chicken) and humans.

Parasites are a neglected area in genomic research, and publicly available reference genomes are often of low quality and riddled with contaminations. Removing such contaminations was an essential part of our workflow, which eventually lead to the identification of the most abundant parasites in chicken (i.e.

Eimeria spp.) and pig (i.e. Blastocystis spp.). Using compositional data analysis, we showed that the bacterial genus Prevotella is positively correlated with the presence of Blastocystis in pig fecal samples. In human fecal samples, we detected Blastocystis subtype 4 predominantly in high-income countries.

Generally, the prevalence of Blastocystis was higher in low-income countries. In the third study, we examined the possibilities of obtaining whole plasmid genomes using metagenomic sequencing. Sewage samples from 22 countries worldwide were sequenced and analyzed to discover potential new plasmids. Using random displacement amplification followed by long-read nanopore sequencing, we were able to detect thousands of circular elements, each reconstructed from individual nanopore reads consisting of multiple tandem repeats of the plasmid sequence.

All three studies contributed to the overall aim of the thesis project: attaining a better understanding of how to apply metagenomics to a diverse set of samples as well as how to extract reliable information on low-abundance entities such as pathogens, parasites, and plasmids.