Benchmarking different approaches for Norovirus genome assembly in metagenome samples

Abstract

Background Genome assembly of viruses with high mutation rates, such as Norovirus and other RNA viruses, or from metagenome samples, poses a challenge for the scientific community due to the coexistence of several viral quasispecies and strains. Furthermore, there is no standard method for obtaining whole-genome sequences in non-related patients. After polyA RNA isolation and sequencing in eight patients with acute gastroenteritis, we evaluated two de Bruijn graph assemblers (SPAdes and MEGAHIT), combined with four different and common pre-assembly strategies, and compared those yielding whole genome Norovirus contigs. Results Reference-genome guided strategies with both host and target virus did not present any advantages compared to the assembly of non-filtered data in the case of SPAdes, and in the case of MEGAHIT, only host genome filtering presented improvements. MEGAHIT performed better than SPAdes in most samples, reaching complete genome sequences in most of them for all the strategies employed. Read binning with CD-HIT improved assembly when paired with different analysis strategies, and more notably in the case of SPAdes. Conclusions Not all metagenome assemblies are equal and the choice in the workflow depends on the species studied and the prior steps to analysis. We may need different approaches even for samples treated equally due to the presence of high intra host variability. We tested and compared different workflows for the accurate assembly of Norovirus genomes and established their assembly capacities for this purpose.

Benchmarking state-of-the-art approaches for norovirus genome assembly in metagenome sample

“…Moreover, the results clearly show that one do not need any additional pre- and post-processing steps beyond quality trimming to obtain good results contrasting with the assembly pipeline presented in (Fuentes-Trillo et al, 2021) that included multiple steps of read binning, contamination filtering and norovirus read filtering. All these steps might influence the final result and cause the degradation of assembly quality in general.…”

“…The results obtained are summarized in Table 1. They clearly show that rnaviral SPA des and corona SPA des (version of rnaviral SPA des that could use profile HMM models to guide an assembly) are better suited for assembly of this data than rnaSPAdes as well as metaSPAdes and MEGAHIT (these two were benchmarked by Fuentes-Trillo et al (2021)). We note that each sample was assembled into a single contig with perfect or near-perfect quality.…”

“…Since our approach only includes read trimming as a preprocessing step and there-fore can be directly compared to pC approach of (Fuentes-Trillo et al, 2021). Here we see that coronaSPAdes was able to assemble 7 out of 8 samples into a contig longer than 7,500 bp, and the contig length of the remaining sample is 7,493 bp that places this sample into a near-complete category.…”

“…Recently, an article by Fuentes-Trillo et al (2021) discussing various approaches for Norovirus genome assembly was published in BMC Genomics. Using the approaches presented in this article we want to highlight some common benchmarking problems that might result in misleading conclusions and that can be easily avoided.…”

“…

Motivation: A recently published article in BMC Genomics by Fuentes-Trillo et al (2021) contains a comparison of assembly approaches of several Noroviral samples via different tools and preprocessing strategies. Unfortunately the study used outdated versions of tools as well as tools that were not designed for the viral assembly task.

…”