Accurate, multi-kb reads resolve complex populations and detect rare microorganisms

Abstract

Accurate evaluation of microbial communities is essential for understanding global biogeochemical processes and can guide bioremediation and medical treatments. Metagenomics is most commonly used to analyze microbial diversity and metabolic potential, but assemblies of the short reads generated by current sequencing platforms may fail to recover heterogeneous strain populations and rare organisms. Here we used short (150-bp) and long (multi-kb) synthetic reads to evaluate strain heterogeneity and study microorganisms at low abundance in complex microbial communities from terrestrial sediments. The long-read data revealed multiple (probably dozens of) closely related species and strains from previously undescribed Deltaproteobacteria and Aminicenantes (candidate phylum OP8). Notably, these are the most abundant organisms in the communities, yet short-read assemblies achieved only partial genome coverage, mostly in the form of short scaffolds (N50 = ∼2200 bp). Genome architecture and metabolic potential for these lineages were reconstructed using a new synteny-based method. Analysis of long-read data also revealed thousands of species whose abundances were <0.1% in all samples. Most of the organisms in this “long tail” of rare organisms belong to phyla that are also represented by abundant organisms. Genes encoding glycosyl hydrolases are significantly more abundant than expected in rare genomes, suggesting that rare species may augment the capability for carbon turnover and confer resilience to changing environmental conditions. Overall, the study showed that a diversity of closely related strains and rare organisms account for a major portion of the communities. These are probably common features of many microbial communities and can be effectively studied using a combination of long and short reads.

Aquifer environment selects for microbial species cohorts in sediment and groundwater

“…In addition, our ability to track organisms through a series of data sets with stringent read-mapping would be hampered by the universally highly conserved regions in the 16 S rRNA gene, whereas for ribosomal protein-encoding scaffolds, the level of divergence is more consistent across the full span of the sequence used. As for the selection of nucleotide identity to define taxonomic units tracked in this study, Konstantinidis et al (2006) showed that strains could robustly be defined based on 498% ANI, with species thus bounded by 95-98% ANI, whereas our previous work identified 98% ANI as a species boundary (Sharon et al, 2015). Hence, we have clustered organisms at 98% global sequence identity as a conservative definition of species, and as a threshold allowing robust mapping of reads for abundance estimates.…”

“…Species predictions were based on a prior analysis of ribosomal protein S3 (rpS3) divergence, which identified 98% and 90% nucleotide identities as thresholds for species and genera, respectively (Sharon et al, 2015). Expanding to genera, only an additional 10 organisms from the two data sets would be considered shared across the communities.…”

“…The rpS3 identities tally closely with sequence identities between global alignments of scaffolds encoding the ribosomal protein block when identities are 490% (R 2 ¼ 0.82), meaning the rpS3 divergence is indicative of the average orthologous gene identity between relatively closely related genomes (Supplementary Table S2). The majority of organisms in the two communities were less closely related, with 64% of the organisms sharing o70% rpS3 nucleotide identity with any member of the other sample's community, a threshold delineating family-level divergence or greater (Sharon et al, 2015). The two sites are separated by B50 meters and sampling occurred 4 years apart, so the community divergence may be a function of spatial and/or temporal separation, of altered geochemical conditions, or some combination of the three.…”

Read Clouds Uncover Variation in Complex Regions of the Human Genome

“…Aligning with RFA may enable determination of the correct haplotype paths of these variable regions. Our approach may also prove useful for subspecies discovery and quantification in metagenomic samples for which previous work had much higher sequencing requirements using TruSeq-assembled synthetic long reads (Sharon et al 2015). There has been previous work to leverage PacBio sequencing to accurately resolve RNA transcripts (Sharon et al 2013).…”

Haplotype-phased synthetic long reads from short-read sequencing

“…However, the loss of linkage information during the generation of short reads limits their utility. In particular, short reads are insufficient to phase the haplotypes of individuals within mixtures of similar sequences, including homeologous and homologous chromosomes in polyploids [3,4], viral quasispecies [5], multiply or alternatively spliced mRNA [6], genes from metagenomic samples containing related organisms [7,8], and immune antibody gene repertoires [9]. In these cases, additional information is required to determine whether mutations separated by distances longer than the read length are present in the same individual.…”