Metagenome Fragment Classification Using N‐Mer Frequency Profiles

Abstract

A vast amount of microbial sequencing data is being generated through large-scale projects in ecology, agriculture, and human health. Efficient high-throughput methods are needed to analyze the mass amounts of metagenomic data, all DNA present in an environmental sample. A major obstacle in metagenomics is the inability to obtain accuracy using technology that yields short reads. We construct the unique N-mer frequency profiles of 635 microbial genomes publicly available as of February 2008. These profiles are used to train a naive Bayes classifier (NBC) that can be used to identify the genome of any fragment. We show that our method is comparable to BLAST for small 25 bp fragments but does not have the ambiguity of BLAST's tied top scores. We demonstrate that this approach is scalable to identify any fragment from hundreds of genomes. It also performs quite well at the strain, species, and genera levels and achieves strain resolution despite classifying ubiquitous genomic fragments (gene and nongene regions). Cross-validation analysis demonstrates that species-accuracy achieves 90% for highly-represented species containing an average of 8 strains. We demonstrate that such a tool can be used on the Sargasso Sea dataset, and our analysis shows that NBC can be further enhanced.

Centrifuge: rapid and sensitive classification of metagenomic sequences

“…Most of the current metagenomics classification programs either suffer from slow classification speed, a large index size, or both. For example, machine-learning-based approaches such as the Naive Bayes Classifier (NBC) (Rosen et al 2008) and PhymmBL Salzberg 2009, 2011) classify <100 reads per minute, which is too slow for data sets that contain millions of reads. In contrast, the pseudoalignment approach employed in Kraken (Wood and Salzberg 2014) processes reads far more quickly, more than 1 million reads per minute, but its exact k-mer matching algorithm requires a large index.…”

Neural network-based taxonomic clustering for metagenomics

“…This classifier was first implemented for organism classification by Sandberg in 2001 on a small set of just 28 genomes, and has since been further extended to a larger database of 635 genomes by Rosen [9,16]. The outputted scores for each fragment are then submitted as features to an unsupervised clustering algorithm.…”

“…Most supervised classification methods for metagenomics employ either a homology-based alignment or a composition-based frequency model [8,9,10,11]. However, as mentioned above, either of these techniques can only identify the 1-2% of organisms (those that are known), and perhaps classify another 50-70% to a higher taxonomic level (such as order or phylum).…”

A Two-Way Bayesian Mixture Model for Clustering in Metagenomics

“…Most of the existing clustering methods are supervised and depend on the availability of reference data for training [15,3,19,5]. A metagenome may however, contain reads from unexplored phyla which cannot be labeled into one of the existing classes.…”

“…The dominant patterns in the data are captured by its component distributions. Most mixture models assume an underlying normal distribution [19]. However, the distribution of word counts within a genome vary according to a Poisson distribution [17,18].…”