Minimum entropy decomposition: unsupervised oligotyping for sensitive partitioning of high-throughput marker gene sequences.

Eren AM, Morrison HG, Lescault PJ, Reveillaud J, Vineis JH, Sogin ML. · 2015

Read Minimum entropy decomposition: unsupervised oligotyping for sensitive partitioning of high-throughput marker gene sequences. by Eren AM, Morrison HG, Lescault PJ, Reveillaud J, Vineis JH, Sogin ML. (2015) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4817710.txt.

Introduction

A. Murat Eren leads six authors in The ISME Journal in 2015 with a sharper way to sort microbe sequences. Surveys of a community read a marker gene from every member, then cluster near-identical reads into units. Fixed similarity cutoffs blur close kin, even when a single letter separates species that live in different niches.

Their method, minimum entropy decomposition, splits the data only at the letter positions that carry real information, a measure borrowed from information theory. Applied to sponges from the deep sea, it split one broad cluster into units faithful to each sponge species. On human mouth data it separated kinds differing by under one percent that nonetheless kept to distinct mouth sites. The method needs no hand tuning and little computing.

Catalog record

CREATOREren AM, Morrison HG, Lescault PJ, Reveillaud J, Vineis JH, Sogin ML.
YEAR2015
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC4817710

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Minimum entropy decomposition: Unsupervised oligotyping for sensitive partitioning of high-throughput marker gene sequences By A Murat Eren, Hilary G Morrison, Pamela J Lescault, Julie Reveillaud, Joseph H Vineis, Mitchell L Sogin Published in: The ISME Journal DOI: 10.1038/ismej.2014.195 ABSTRACT Molecular microbial ecology investigations often employ large marker gene datasets, for example, ribosomal RNAs, to represent the occurrence of single-cell genomes in microbial communities. Massively parallel DNA sequencing technologies enable extensive surveys of marker gene libraries that sometimes include nearly identical sequences. Computational approaches that rely on pairwise sequence alignments for similarity assessment and de novo clustering with de facto similarity thresholds to partition high-throughput sequencing datasets constrain fine-scale resolution descriptions of microbial communities. Minimum Entropy Decomposition (MED) provides a computationally efficient means to partition marker gene datasets into ‘MED nodes', which represent homogeneous operational taxonomic units. By employing Shannon entropy, MED uses only the information-rich nucleotide positions across reads and iteratively partitions large datasets while omitting stochastic variation. When applied to analyses of microbiomes from two deep-sea cryptic sponges Hexadella dedritifera and Hexadella cf. dedritifera, MED resolved a key Gammaproteobacteria cluster into multiple MED nodes that are specific to different sponges, and revealed that these closely related sympatric sponge species maintain distinct microbial communities. MED analysis of a previously published human oral microbiome dataset also revealed that taxa separated by less than 1% sequence variation distributed to distinct niches in the oral cavity. The information theory-guided decomposition process behind the MED algorithm enables sensitive discrimination of closely related organisms in marker gene amplicon datasets without relying on extensive computational heuristics and user supervision. INTRODUCTION Marker gene analyses of microbial diversity require categorizing DNA sequences into ecologically meaningful units. The two major approaches for partitioning large datasets include: (i) taxonomic classification of sequences through comparison with curated databases, for example, GreenGenes (DeSantis et al., 2006; McDonald et al., 2012) or SILVA (Pruesse et al., 2007; Quast et al., 2013) and (ii) de novo clustering by sequence similarity to define operational taxonomic units (OTUs). The number of unique taxonomic assignments in reference databases limits diversity descriptions because microbiologists have not defined a unified species concept (Gevers et al., 2005; Doolittle and Zhaxybayeva, 2009) and molecular databases generally lack discrete name assignments for the astonishing number of uncultured microbes (Stewart, 2012). In contrast, taxonomy-independent OTU clustering divides datasets into OTUs without requiring a curated taxonomic database (Sun et al., 2012). Database independence frees OTU clustering approaches from the apparent limitations of taxonomy and allows the detection of organisms that have not yet been characterized (Ley et al., 2006). FastGroup, among the first published algorithm to assign sequences to OTUs, described the three major steps employed by most contemporary OTU finding algorithms: ‘(1) compare all the sequences in a dataset to each other, (2) group similar sequences (>=97% identical to each other) together, and (3) output a representative sequence from each group' (Seguritan and Rohwer, 2001). During the last 10 years, various de novo approaches have facilitated OTU identification, including algorithms that typically use pairwise alignments to compute a distance matrix before grouping sequences into OTUs (DOTUR (Schloss and Handelsman, 2005), ESPRIT (Sun et al., 2009), SLP (Huse et al., 2010), HPC-CLUST (Matias Rodrigues and von Mering, 2014)) as well as greedy but more computationally efficient heuristics that perform sequence comparison and OTU identification simultaneously (i.e., ESPRIT-Tree (Cai and Sun, 2011), CD-HIT (Li et al., 2001), UCLUST (Edgar, 2010), DySC (Zheng et al., 2012)). Various software platforms, including mothur (Schloss et al., 2009), QIIME (Caporaso et al., 2010), CD-HIT Suite (Huang et al., 2010) and VAMPS (Huse et al., 2014), have adopted most of these OTU identification strategies. Pairwise sequence identity computes the number of mismatched nucleotides between two aligned reads. However, random PCR or sequencing errors reduce the similarity between two sequences and result in false OTUs even after stringent quality filtering (Huse et al., 2007; Quince et al., 2009; Eren et al., 2013b). To lessen the impact of sequencing errors and subsequent OTU inflation, researchers use relaxed identity thresholds for de novo clustering. For example, the de facto 97% identity often defines the diversity of clusters for 16S rRNA data. Although it reduces the number of observed OTUs, it also generates phylogenetically mixed OTUs (Koeppel and Wu, 2013). Such OTUs aggregate distinct organisms and conceal ecologically important findings (Eren et al., 2013a). Not all nucleotide positions in a dataset contribute equally towards partitioning marker gene data into ecologically meaningful units. Woese (Woese et al., 1985) used short, informative oligonucleotide signatures in the 16S rRNA gene to distinguish between major bacterial clades. As recently described, oligotyping (Eren et al., 2013a) also employs a form of signature analysis by using Shannon entropy (Shannon, 1948) to distinguish biologically meaningful signals from noise without requiring the calculation of pairwise sequence similarity. By relying on information-rich variable sites and discarding low-entropy nucleotide positions in a group of sequencing reads, oligotyping facilitates the identification of closely related but distinct organisms that may differ by as little as one nucleotide over the sequenced region of the 16S rRNA gene. Oligotyping different high-throughput sequencing datasets has resolved unexplained diversity within taxa and OTUs (Eren et al., 2011; Mclellan et al., 2013; Apprill et al., 2014; Eren et al., 2014; Maignien et al., 2014; Reveillaud et al., 2014). However, oligotyping has characteristics that sometimes limit its applicability to environmental datasets. First, oligotyping performs optimally when applied to closely related taxa as the great number of high-entropy locations among distantly related organisms renders the supervision step arduous (Eren et al., 2013a). Second, oligotyping requires a preliminary OTU clustering or taxonomic classification to identify closely related taxa suitable for analysis. Oligotyping is not a stand-alone approach and its improvements cannot be applied directly to the entirety of a sequencing dataset. Nevertheless, oligotyping shows the potential of entropy to partition mixed information into homogenous units by using a fraction of available sequencing data, and without pairwise sequence alignment and comparison (Figure 1). We have developed and employed a new algorithm, ‘Minimum Entropy Decomposition' (MED), to partition high-throughput sequencing datasets into ecologically meaningful and phylogenetically homogeneous units by extending the underlying principles of oligotyping to entire marker gene datasets. MED uses information uncertainty among sequence reads to iteratively decompose a dataset until the maximum entropy criterion is satisfied for each final unit (i.e., until there is no entropy left to explain; Figure 2). In contrast to oligotyping, MED requires no user supervision, no preliminary classification or clustering result, and can be applied to the entire dataset instead of only a group of closely related taxa. We used two datasets to demonstrate the utility of MED: a new V4-V5 rRNA gene dataset from previously described host microbiomes of deep-sea sponges (Reveillaud et al., 2014) and a publicly available human oral microbiome dataset (The Human Microbiome Project Consortium, 2012b). We compared MED analysis results to taxonomic analysis and de novo OTU clustering using QIIME with UCLUST, which among several methods, has often been used in studies of the Human Microbiome Project (The Human Microbiome Project Consortium, 2012a) and the Earth Microbiome Project (Gilbert et al., 2010). MATERIALS AND METHODS Minimum Entropy Decomposition The algorithm iteratively partitions a dataset of amplicon sequences into homogenous OTUs (‘MED nodes') that

Read the complete text