Discovering the Unknown: Improving Detection of Novel Species and Genera from Short Reads

Abstract

High-throughput sequencing technologies enable metagenome profiling, simultaneous sequencing of multiple microbial species present within an environmental sample. Since metagenomic data includes sequence fragments (“reads”) from organisms that are absent from any database, new algorithms must be developed for the identification and annotation of novel sequence fragments. Homology-based techniques have been modified to detect novel species and genera, but, composition-based methods, have not been adapted. We develop a detection technique that can discriminate between “known” and “unknown” taxa, which can be used with composition-based methods, as well as a hybrid method. Unlike previous studies, we rigorously evaluate all algorithms for their ability to detect novel taxa. First, we show that the integration of a detector with a composition-based method performs significantly better than homology-based methods for the detection of novel species and genera, with best performance at finer taxonomic resolutions. Most importantly, we evaluate all the algorithms by introducing an “unknown” class and show that the modified version of PhymmBL has similar or better overall classification performance than the other modified algorithms, especially for the species-level and ultrashort reads. Finally, we evaluate the performance of several algorithms on a real acid mine drainage dataset.

Rapid and accurate taxonomic classification of insect (class Insecta) cytochrome c oxidase subunit 1 (COI) DNA barcode sequences using a naïve Bayesian classifier

“…An alternative approach uses a ‘detector’ to flag sequences that represent taxa not present in the reference data set so that they can be removed prior to making taxonomic assignments (Rosen et al . ; Lan et al . ).…”

Consistency of metagenomic assignment programs in simulated and real data

“…However, at least two issues make taxonomic assignment difficult. First, the read length obtained by next generation sequencing technologies is not long enough to allow the original methods to properly assign the reads to low taxonomic levels (such as genus or species) due to the low sequence divergence between closely related taxonomic groups [ 3 ]. And second, since reference genomes are not available for many uncultured organisms, an incorrect assignment (or even no assignment at all) may be produced when no closely related species have been previously identified.…”

“…Packages using the former approach include MG-RAST [ 4 ] and MEGAN [ 5 ], while the Naïve Bayesian Classifier (such as implemented in Fragment Classification Package, FCP) [ 6 ] and the interpolated Markov model classification (IMM-based), used by Phymm [ 7 ] are based on composition similarity. The performance of assignment programs has been assessed using simulated and well-known experimental data [ 2 , 3 , 6 - 8 ]. In the case of composition-based programs, these methods can classify all the reads [ 2 ], and report the associated likelihood of the read to be assigned to the different categories.…”

Direct 16S rRNA-seq from bacterial communities: a PCR-independent approach to simultaneously assess microbial diversity and functional activity potential of each taxon

“…While new sequences are regularly added to databases (more than ten million 16S-rRNA gene sequences are currently available, source: ncbi.nlm.nih.gov), a series of intrinsic limitations affecting the different experimental protocols have been highlighted 22 23 24 25 26 27 28 . These aspects prompted both protocol improvements 29 and suggestions for denoising or data mining 30 31 32 33 .…”