Metagenomic assembly is the main bottleneck in the identification of mobile genetic elements

Abstract

Antimicrobial resistance genes (ARG) are commonly found on acquired mobile genetic elements (MGEs) such as plasmids or transposons. Understanding the spread of resistance genes associated with mobile elements (mARGs) across different hosts and environments requires linking ARGs to the existing mobile reservoir within bacterial communities. However, reconstructing mARGs in metagenomic data from diverse ecosystems poses computational challenges, including genome fragment reconstruction (assembly), high-throughput annotation of MGEs, and identification of their association with ARGs. Recently, several bioinformatics tools have been developed to identify assembled fragments of plasmids, phages, and insertion sequence (IS) elements in metagenomic data. These methods can help in understanding the dissemination of mARGs. To streamline the process of identifying mARGs in multiple samples, we combined these tools in an automated high-throughput open-source pipeline, MetaMobilePicker, that identifies ARGs associated with plasmids, IS elements and phages, starting from short metagenomic sequencing reads. This pipeline was used to identify these three elements on a simplified simulated metagenome dataset, comprising whole genome sequences from seven clinically relevant bacterial species containing 55 ARGs, nine plasmids and five phages. The results demonstrated moderate precision for the identification of plasmids (0.57) and phages (0.71), and moderate sensitivity of identification of IS elements (0.58) and ARGs (0.70). In this study, we aim to assess the main causes of this moderate performance of the MGE prediction tools in a comprehensive manner. We conducted a systematic benchmark, considering metagenomic read coverage, contig length cutoffs and investigating the performance of the classification algorithms. Our analysis revealed that the metagenomic assembly process is the primary bottleneck when linking ARGs to identified MGEs in short-read metagenomics sequencing experiments rather than ARGs and MGEs identification by the different tools.

Keywords: Bacteria; Bacterial; Computational biology; Drug resistance; Metagenomics; Plasmid.

Figure 1. Workflow of the MetaMobilePicker pipeline.

Colors indicate the different steps/modules in the pipeline. Light blue: preprocessing and assembly. Blue: MGE identification. Dark blue: ARG annotation. Orange: output construction. Software tools used are indicated in brackets. AMR: Antimicrobial resistance. MGE: Mobile Genetic Element. IS: Insertion Sequence. The pipeline is available at http://metamobilepicker.nl. For software references and versions see Table S1.

Figure 2. Annotation plot of bacterial reference genomes in validation dataset.

Blue rectangles: insertion sequences. Orange arrows: antimicrobial resistance genes. (A) Klebsiella pneumoniae chromosome and three plasmids. (B) Acinetobacter baumannii chromosome and plasmid. (C) Enterococcus faecalis chromosome and plasmid. (D) Mycobacterium tuberculosis chromosome. (E) Salmonella enterica chromosome and plasmid. (F) Staphylococcus aureus chromosome. (G) Escherichia coli chromosome and plasmid.

Figure 3. Percentage of reads covering the reference genomes per sampled depth, averaged per species.

Solid lines indicate bacterial origin, dotted lines indicate phage origin.

Figure 4. Circular graph displaying the mapping of antibiotic resistance genes (ARGs) (inner lines) on assembled contigs (light blue inner circle) and reference genomes (dark blue inner circle).

Second circle displays sequence origin and classification. Purple: (true positive) chromosomes. Green: (true positive) plasmids. Blue: (true positive) phages. Red: False positive plasmids. Yellow: false negative plasmids. Orange lines in the outer ring denote the location of the ARGs. Outer lines denote ARGs present more than once in the reference genomes.

Figure 5. Number of false positive plasmid predictions per replicon.