imGLAD: Accurate detection and quantification of target organisms in metagenomes

Abstract

Accurate detection of target microbial species in metagenomic datasets from environmental samples remains limited because the limit of detection of current methods is typically inaccessible and the frequency of false-positives, resulting from inadequate identification of regions of the genome that are either too highly conserved to be diagnostic (e.g., rRNA genes) or prone to frequent horizontal genetic exchange (e.g., mobile elements) remains unknown. To overcome these limitations, we introduce imGLAD, which aims to detect genomic sequences in metagenomic datasets. imGLAD achieves high accuracy because it uses the sequence-discrete population concept for discriminating between metagenomic reads originating from the target organism compared to reads from co-occurring close relatives, masks regions of the genome that are not informative using the MyTaxa engine, and models both the sequencing breadth and depth to determine relative abundance and limit of detection. We validated imGLAD by analysing metagenomic datasets derived from spinach leafs inoculated with the enteric pathogen Escherichia coli O157:H7 and showed that its limit of detection is comparable to that of PCR-based approaches (~1 cell/gram). ABSTRACTAccurate detection of target microbial species in metagenomic datasets from environmental samples remains limited because the limit of detection of current methods is typically inaccessible and the frequency of false-positives, resulting from inadequate identification of regions of the genome that are either too highly conserved to be diagnostic (e.g., rRNA genes) or prone to frequent horizontal genetic exchange (e.g., mobile elements) remains unknown. To overcome these limitations, we introduce imGLAD, which aims to detect genomic sequences in metagenomic datasets. imGLAD achieves high accuracy because it uses the sequence-discrete population concept for discriminating between metagenomic reads originating from the target organism compared to reads from co-occurring close relatives, masks regions of the genome that are not informative using the MyTaxa engine, and models both the sequencing breadth and depth to determine relative abundance and limit of detection. We validated imGLAD by analysing metagenomic datasets derived from spinach leafs inoculated with the enteric pathogen Escherichia coli O157:H7 and showed that its limit of detection is comparable to that of PCR-based approaches (~1 cell/gram).

Iterative subtractive binning of freshwater chronoseries metagenomes identifies over 400 novel species and their ecologic preferences

“…4). WB species were considered present in a sample if their sequencing breadth was at least 10%, which corresponds to confidence of presence >95% (Castro et al ., 2018). In order to determine environmental or geographic preference, we estimated preference scores based on the frequencies of presence in different sets of samples, normalizing by the baseline distribution of each gsp and the probability of capturing any gsp in a given sample by chance, and implicitly accounting for sample size and community evenness among other factors (see Materials and methods section; Fig.…”

“…A gsp was considered to be present in a sample if the TAD 80 was non‐zero. Note that non‐zero TAD 80 is equivalent to sequencing breadth ≥10% (i.e., at least 10% of genome positions covered by sequencing reads), previously shown to correspond to confidence of species presence >95% (Castro et al ., 2018). The alpha diversity was estimated using the Nonpareil sequence diversity index ( N d ) projected to Shannon diversity H′ (Nonpareil) (Rodriguez‐R, Gunturu, Tiedje, et al ., 2018c), as well as H′ on the gspp abundance profile (AlphaDiversity.pl) (Rodriguez‐R and Konstantinidis, 2016).…”

Iterative Subtractive Binning of Freshwater Chronoseries Metagenomes Identifies of over Four Hundred Novel Species and their Ecologic Preferences

“…4). WB species were considered present in a sample if their sequencing depth was at least 10%, which corresponds to confidence of presence > 95% [51]. In order to determine environmental or geographic preference, we estimated preference scores based on the frequencies of presence in different sets of samples, normalizing by the baseline distribution of each gsp and the probability of capturing any gsp in a given sample, and implicitly accounting for sample size and community evenness among other factors (see Methods; Fig.…”

“…In order to evaluate preference (geographic or environmental), we devised a metric to compare expected and observed presence frequencies (Fig. 5) based on the observation that “presence” can be confidently assessed at the species level (95% ANI) and 0.05 p-value significance given a genome sequencing breadth of at least 10% [51]. All detected species appeared to have a preference for either freshwater or saltwater, or were too scarcely present to determine preference.…”

“…Abundance was estimated as TAD normalized by the genome equivalents of the metagenomic dataset (MicrobeCensus [50]), resulting in units of community fraction. A gsp was considered to be present in a sample if the TAD was non-zero (equivalent to sequencing breadth ≥ 10%, previously shown to correspond to confidence of presence > 95% [51]). The alpha-diversity was estimated using the sequence diversity N d projected to Shannon diversity H’ (Nonpareil [3]), as well as H’ on the gspp abundance profile (AlphaDiversity.pl [21]).…”