Identifying contamination with advanced visualization and analysis practices: metagenomic approaches for eukaryotic genome assemblies

Abstract

High-throughput sequencing provides a fast and cost-effective mean to recover genomes of organisms from all domains of life. However, adequate curation of the assembly results against potential contamination of non-target organisms requires advanced bioinformatics approaches and practices. Here, we re-analyzed the sequencing data generated for the tardigrade Hypsibius dujardini, and created a holistic display of the eukaryotic genome assembly using DNA data originating from two groups and eleven sequencing libraries. By using bacterial single-copy genes, k-mer frequencies, and coverage values of scaffolds we could identify and characterize multiple near-complete bacterial genomes from the raw assembly, and curate a 182 Mbp draft genome for H. dujardini supported by RNA-Seq data. Our results indicate that most contaminant scaffolds were assembled from Moleculo long-read libraries, and most of these contaminants have differed between library preparations. Our re-analysis shows that visualization and curation of eukaryotic genome assemblies can benefit from tools designed to address the needs of today's microbiologists, who are constantly challenged by the difficulties associated with the identification of distinct microbial genomes in complex environmental metagenomes.

Keywords: Assembly; Contamination; Curation; Genomics; HGT; Visualization.

Figure 1. Holistic assessment of the tardigrade genome assembly from Boothby et al. (2015).

Dendrogram in the center organizes scaffolds based on sequence composition, and coverage values acquired from 11 DNA libraries. Scaffolds larger than 40 kbp were split into sections of 20 kbp for visualization purposes. Splits are displayed in the first inner circle and GC-content (0–71%) in the second circle. In the following 11 layers, each bar represents the portion of scaffolds covered by short reads in a given sample. The next layer shows the same information for RNA-Seq data. Scaffolds harboring genes used by Boothby et al. to support the expended HGT hypothesis is shown in the next layer. Finally, the outermost layer shows our selections of scaffolds as draft genome bins: the curated tardigrade genome (selection #1), as well as three near-complete bacterial genomes originating from various contamination sources (selections #2, #3, and #4).

Figure 2. Occurrence of the 139 bacterial single-copy genes reported by Campbell et al. (2013) across scaffold collections.

The top two plots display the frequency and distribution of single-copy genes in the raw tardigrade genomic assembly generated by Boothby et al. (2015), and Koutsovoulos et al. (2016), respectively. The bottom two plots display the same information for each of the curated tardigrade genomes. Each bar represents the squared-root normalized number of significant hits per single-copy gene. The same information is visualized as box-plots on the left side of each plot.