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

Abstract

Background Over the past years, sequencing technologies have expanded our ability to examine novel microbial metabolisms and diversity previously obscured by isolation approaches. Long-read sequencing promises to revolutionize the metagenomic field and recover less fragmented genomes from environmental samples. Nonetheless, how to best benefit from long-read sequencing and whether long-read sequencing can provide recovered genomes of similar characteristics as short-read approaches remains unclear. Results We recovered metagenome-assembled genomes (MAGs) from the free-living fraction at four-time points during a spring bloom in the North Sea. The taxonomic composition of all MAGs recovered was comparable between technologies. However, differences consisted of higher sequencing depth for contigs and higher genome population diversity in short-read compared to long-read metagenomes. When pairing population genomes recovered from both sequencing approaches that shared ≥ 99% average nucleotide identity, long-read MAGs were composed of fewer contigs, a higher N50, and a higher number of predicted genes when compared to short-read MAGs. Moreover, 88% of the total long-read MAGs carried a 16S rRNA gene compared to only 23% of MAGs recovered from short-read metagenomes. Relative abundances for population genomes recovered using both technologies were similar, although disagreements were observed for high and low GC content MAGs. Conclusions Our results highlight that short-read technologies recovered more MAGs and a higher number of species than long-read due to an overall higher sequencing depth. Long-read samples produced higher quality MAGs and similar species composition compared to short-read sequencing. Differences in the GC content recovered by each sequencing technology resulted in divergences in the diversity recovered and relative abundance of MAGs within the GC content boundaries.

BASALT refines binning from metagenomic data and increases resolution of genome-resolved metagenomic analysis

“…As Aiding Lake sediment samples only contained SRS, we further supplemented nine datasets, including four datasets with both SRS and LRS, and five PacBio High-Fidelity (HiFi) datasets, to assess the performance of BASALT on real samples. The four SRS + LRS datasets comprised a subset dataset from human gut microbiome (ten Illumina SRS samples, 204 GB in total, ten Oxford Nanopore (ONT) LRS samples, 113.6 GB in total) 42 , a subset dataset from marine plankton microbiome (four Illumina SRS samples, 263.8 GB in total, and four Pacbio LRS samples, 91.6 GB in total) 43 , a dataset from activated sludge microbiome (two Illumina SRS samples, 245.6 GB in total, and three ONT samples, 105.8 GB in total) 44 , and a dataset from Antarctic soil microbiome (one Illumina sample 67.2 GB, one ONT sample 83.5 GB) 45 . The five PacBio HiFi datasets comprised a human gut microbiome (five samples, 182.6 GB in total) 46 , a sheep gut microbiome (one sample, 92.1 GB) 47 , a chicken gut microbiome (three samples, 366.8 GB in total) 48 , a hot spring sediment microbiome (one sample 53.2 GB) 49 , and an anaerobic digester microbiome (one sample 28.6 GB) 50 .…”

Uncultured prokaryotic genomes in the spotlight: An examination of publicly available data from metagenomics and single-cell genomics

“…It is challenging to construct multiple species-resolved MAGs from complex prokaryotic populations because most contigs are unassigned and discarded in the binning process [88] . The use of long-read sequencing technologies, such as PacBio [121] , [122] , [123] and Oxford Nanopore Technologies [124] , [125] , [126] , [127] , can help overcome these issues. Future developments in binning algorithms that leverage machine learning [23] , [128] and Hi-C metagenomics [121] , [129] , [130] may help address these challenges.…”

Implementing Whole Genome Sequencing (WGS) in Clinical Practice: Advantages, Challenges, and Future Perspectives

“…However, the drawback of long-read sequencing is the occurrence of errors [ 105 , 106 ]. A great comparison can be found in a recently published metagenomic study, in which the authors emphasize the advantages and disadvantages of these two approaches [ 109 ].…”