Benchmarking second and third-generation sequencing platforms for microbial metagenomics

Abstract

Shotgun metagenomic sequencing is a common approach for studying the taxonomic diversity and metabolic potential of complex microbial communities. Current methods primarily use second generation short read sequencing, yet advances in third generation long read technologies provide opportunities to overcome some of the limitations of short read sequencing. Here, we compared seven platforms, encompassing second generation sequencers (Illumina HiSeq 300, MGI DNBSEQ-G400 and DNBSEQ-T7, ThermoFisher Ion GeneStudio S5 and Ion Proton P1) and third generation sequencers (Oxford Nanopore Technologies MinION R9 and Pacific Biosciences Sequel II). We constructed three uneven synthetic microbial communities composed of up to 87 genomic microbial strains DNAs per mock, spanning 29 bacterial and archaeal phyla, and representing the most complex and diverse synthetic communities used for sequencing technology comparisons. Our results demonstrate that third generation sequencing have advantages over second generation platforms in analyzing complex microbial communities, but require careful sequencing library preparation for optimal quantitative metagenomic analysis. Our sequencing data also provides a valuable resource for testing and benchmarking bioinformatics software for metagenomics.

Comparing genomes recovered from time-series metagenomes using long- and short-read sequencing technologies

“…The Illumina SR approach recovered an average of 36 Gbp per sample (~ 240 bp paired-end reads), while the CCS PacBio LR approach recovered an average of 12 Gbp per sample (~ 6 kbp CCS reads) (Table S 1 ). This difference translated into SR metagenomic samples having, on average, ~3 times more sequenced base pairs than LR samples, which is within the expected output for single sequenced samples [ 8 ]. Additionally, LR metagenomic samples had slightly higher GC content (0.43) compared to their SR counterpart (0.38) when all reads were considered (Table S 1 ).…”

“…These differences likely reflect the inherent genomic characteristics of the groups enriched within each sequencing technology [ 46 ]. Biases in low and high GC content spaces are recognized for SR technologies [ 47 ], while reports on LR approaches have noted anywhere from minimal GC biases in mock communities [ 8 ] to a higher recovery of high GC content sequences in metagenomes [ 46 , 48 , 49 ]. According to our results, most SR-only species belonged to Bacteroidia , Alphaproteobacteria , and Gammaproteobacteria , whereas for LR-only species, Acidiimicrobiia , and Verrucomicrobiae were the two major classes.…”

Reliability of species detection in 16S microbiome analysis: Comparison of five widely used pipelines and recommendations for a more standardized approach

“…Strains that are well distinguishable on the basis of 16S rRNA genes could be well identifiable at the species level, indicating a high accuracy of the chosen method. Although the Ion Torrent ® platform might produce higher rates of indel errors, especially in homopolymer regions [ 32 ], comparable performance for 16S amplicon sequencing to Illumina-based workflows has been demonstrated [ 33 ]. However, quality filtering steps and subsequent bioinformatic methods should be adapted to the technology used and validated.…”

In Vitro Modelling of Oral Microbial Invasion in the Human Colon

“…For each sample, 1 μg of high-molecular-weight DNA (>10 kbp) was used to build the library. DNA was sheared into 150-bp fragments using an ultrasonicator (Covaris, Woburn, MA), and DNA fragment library construction was performed using the Ion Plus fragment library and Ion Xpress barcode adapters kit (Thermo Fisher Scientific, USA), as previously described ( 57 ). We used an Ion Proton sequencer and Ion GeneStudio S5 prime sequencer to sequence the libraries (Thermo Fisher Scientific, USA), with a minimum of 20 million 150-bp high-quality reads generated per library for luminal, mucosal, and enriched saliva samples.…”