Comprehensive Evaluation of RNA and DNA Viromic Methods Based on Species Richness and Abundance Analyses Using Marmot Rectal Samples

Abstract

Viral metagenomics is the most powerful tool to profile viromic composition for a given sample. Different viromic methods, including amplification-free ones, have been developed, but choosing them for different purposes requires comprehensive benchmarks. Here, we assessed the performance of four routinely used methods, i.e., multiple displacement amplification (MDA), direct metagenomic sequencing (MTG), sequence-independent single-primer amplification (SIA), and metatranscriptomic sequencing (MTT), using marmot rectal samples as the templates spiked with five known viruses of different genome types. The obtained clean data were differently contaminated by host and bacterial genomes, resulting in MDA having the most, with ~72.1%, but MTT had only ~7.5% data, useful for follow-up viromic analysis. MDA showed a broader spectrum with higher efficiency to profile the DNA virome, and MTT captured almost all RNA viruses with extraordinary sensitivity; hence, they are advisable in richness-based viromic studies. MTG was weak in capturing single-stranded DNA viruses, and SIA could detect both RNA and DNA viruses but with high randomness. Due to biases to certain types of viruses, the four methods caused different alterations to species abundance compared to the initial virus composition. SIA and MDA introduced greater stochastic errors to relative abundances of species, genus, and family taxa, whereas the two amplification-free methods were more tolerant toward such errors and thus are recommendable in abundance-based analyses. In addition, genus taxon is a compromising analytic level that ensures technically supported and biologically and/or ecologically meaningful viromic conclusions. IMPORTANCE Viral metagenomics can be roughly divided into species richness-based studies and species abundance-based analyses. Viromic methods with different principles have been developed, but rational selection of these techniques according to different purposes requires comprehensive understanding of their properties. By assessing the four most widely used methods using template samples, we found that multiple displacement amplification (MDA) and metatranscriptomic sequencing (MTT) are advisable for species richness-based viromic studies, as they show excellent efficiency to detect DNA and RNA viruses. Meanwhile, metagenomic sequencing (MTG) and MTT are more compatible with stochastic errors of methods introduced into relative abundance of viromic taxa and hence are rational choices in species abundance-based analyses. This study also highlights that MTG needs to tackle host genome contamination and ameliorate the capacity to detect single-stranded DNA viruses in the future, and the MTT method requires an improvement in bacterial rRNA depletion prior to library preparation.

Keywords: performance comparison; species abundance; species richness; stochastic error; taxonomic rank; viral metagenomics.

FIG 1

Bacterial and host contamination in high-throughput sequencing (HTS) clean data. (A to C) The alignment rates of small subunit (SSU) (A) and large subunit (LSU) (B) rRNA and bacterial marker (BM) (C) of these data were evaluated using ViromeQC 1.0. (D) These HTS data are composed of host genome, bacterial contamination (classified using Kraken2), and unclassified reads. MDA, multiple displacement amplification; MTG, direct metagenomic sequencing; SIA, sequence-independent single-primer amplification; MTT, meta-transcriptomic sequencing.

FIG 2

Virus operational taxonomic unit (vOTU) accumulation curves of MDA (A), MTG (B), SIA (C), and MTT (D) viromic techniques in repeated sequencing (RPS) (referring to the bottom axes) and ultradeep sequencing (UDS) (referring to the top axes).

FIG 3

Principal coordinates analysis (PCoA) analyses revealed that libraries (triangles) aggregated into four clusters corresponding to MDA, MTG, SIA, and MTT viromic techniques that show different preferences to vOTUs (filled circles) of double-stranded DNA (dsDNA) (A), single-stranded DNA (ssDNA) (B), circular single-stranded DNA (cssDNA) (C), reverse transcribing RNA (rtRNA) (D), double-stranded RNA (dsRNA) (E), single-stranded RNA (and ssRNA) (F). The circled libraries are rebuilt MTG and MTT ones that detach from the initial ones due to batch effects.

FIG 4

Stochastic errors of the four methods cause different variation to relative abundance (A) of taxa at the vOTU, genus, and family levels, which further affect Shannon (B) and Simpson (C) α-diversity indices and Bray-Curtis distances (D) between libraries. CV, coefficient of variation; RA, relative abundance. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

FIG 5

The variations in the relative abundance of vOTU, genus, and family taxa are negatively correlated with relative abundance.

FIG 6

Read numbers (left column), the complete genome coverage (middle column), and the linear relationship (right column) between the read abundance and the viral concentration (gene copies/μL) of five spiking viruses in the four viromic techniques. The colors in each panel represent the different viruses, as shown in the key at the bottom of the figure. RPKM, reads per kilobase per million mapped reads; PCV2, porcine circovirus 2; PPV1, porcine parvovirus 1; PRV, porcine pseudorabies virus; RVA, group A rotavirus; RABV, rabies virus.