Large-scale machine learning for metagenomics sequence classification

Abstract

Motivation: Metagenomics characterizes the taxonomic diversity of microbial communities by sequencing DNA directly from an environmental sample. One of the main challenges in metagenomics data analysis is the binning step, where each sequenced read is assigned to a taxonomic clade. Because of the large volume of metagenomics datasets, binning methods need fast and accurate algorithms that can operate with reasonable computing requirements. While standard alignment-based methods provide state-of-the-art performance, compositional approaches that assign a taxonomic class to a DNA read based on the k-mers it contains have the potential to provide faster solutions.Results: We propose a new rank-flexible machine learning-based compositional approach for taxonomic assignment of metagenomics reads and show that it benefits from increasing the number of fragments sampled from reference genome to tune its parameters, up to a coverage of about 10, and from increasing the k-mer size to about 12. Tuning the method involves training machine learning models on about 108 samples in 107 dimensions, which is out of reach of standard softwares but can be done efficiently with modern implementations for large-scale machine learning. The resulting method is competitive in terms of accuracy with well-established alignment and composition-based tools for problems involving a small to moderate number of candidate species and for reasonable amounts of sequencing errors. We show, however, that machine learning-based compositional approaches are still limited in their ability to deal with problems involving a greater number of species and more sensitive to sequencing errors. We finally show that the new method outperforms the state-of-the-art in its ability to classify reads from species of lineage absent from the reference database and confirm that compositional approaches achieve faster prediction times, with a gain of 2–17 times with respect to the BWA-MEM short read mapper, depending on the number of candidate species and the level of sequencing noise.Availability and implementation: Data and codes are available at http://cbio.ensmp.fr/largescalemetagenomics.Contact: pierre.mahe@biomerieux.comSupplementary information: Supplementary data are available at Bioinformatics online.

Machine learning for metagenomics: methods and tools

“…Furthermore, in terms of computation, once a model has been trained using reference data, it's execution on novel sequences is generally linear and thus much less time-consuming than alignment based methods. A recent study by Vervier and co-authors [75] that explores the possibility of a largescale machine learning implementation for taxonomic assignment problem, has con rmed that compositional approaches achieve faster prediction times and consequently are appropriate for whole metagenome studies.…”

High-resolution characterization of the human microbiome

“…LMAT 133 and Kraken 134 both assign taxonomy based on identified LCA taxa for k -mers in each query sequence. Other methods train models on the k -mer profiles associated with each taxon, using a variety of machine learning approaches including neural networks (TAC-ELM 135 ), naïve Bayes classifiers (RITA 136 ), or linear models-based methods 137 . TAC-ELM also incorporates data on GC content and RITA combines BLAST-based reference alignments.…”

“…TAC-ELM also incorporates data on GC content and RITA combines BLAST-based reference alignments. Comparisons between Kraken and the linear model-based method above suggest that while exact k-mer matching methods like LMAT and Kraken are more accurate when query sequences originate from reference genomes, they may produce overly specific classifications for sequences from genomes absent from the reference database 137 . Moreover, Kraken requires fairly long (31 amino acid) k -mer matches, which may potentially reject many short reads due to insufficient data.…”

A Comprehensive Review of Emerging Computational Methods for Gene Identification

“…Recently, machine learning based methods [169,199] are being further used for gene prediction in metagenomic fragments. Also, k-mer-based sequence binning methods and sequence property-based methods are often seen as the input for training models.…”