Comparison of Statistical Methods to Classify Environmental Genomic Fragments

Abstract

"Binning" (or taxonomic classification) of DNA sequence reads is an initial step to analyzing an environmental biological sample. Currently, a homology-based tool, BLAST, is one of the most commonly used tools to label DNA reads, but it is argued that BLAST will quickly lose its classification ability as the genome databases grow. In this paper, we compare the accuracies of a naïve Bayes classifier (NBC) and statistical language model to BLAST for binning reads and demonstrate that NBC obtains good performance for the low cost of computational complexity. On the other hand, the back-off n-gram language model can improve accuracy when only partial training data is available (such as in-progress sequencing projects). NBC demonstrates comparable performance to BLAST and can also be optimized on partial training datasets by adjusting the word feature size. A fivefold cross validation is conducted to compare each method's accuracy for determining novel genomes at different taxonomic levels, with NBC outperforming BLAST for species-level classification but BLAST outperforming NBC for genus-level and phyla-level classification. In conclusion, the NBC is a competitive taxonomic classifier, and language models can improve performance when only partial training data is available.

The transcriptional response of microbial communities in thawing Alaskan permafrost soils

“…Contigs were assigned to high-level taxonomic groups (Class level and above) using a Naïve Baysien Classifier (NBC) that compares against a database containing all DNA sequences in NCBI that classified as either Bacteria, Archaea, Fungi, or viral (Rosen and Essinger, 2010 ). This approach was chosen because NBC has been found to outperform most other composition, similarity, and phylogeny based metagenomic classifiers in terms of sensitivity and precision (Bazinet and Cummings, 2012 ).…”

Statistical Linear Models in Virus Genomic Alignment-free Classification: Application to Hepatitis C Viruses

“…In general, the classifiers performed better at genotyping than at subtyping HCV sequences. Several studies for viral [9], [10] and metagenomic [3], [20], [21] taxonomic classification have reported similar results where the performance is better at high-level classifications. Genomic sequences are more similar at low-level than at high-level clades, which makes more difficult to discriminate between sequences at low-level clades.…”

“…For instance B-MB with α = 1e−100 and LSVM with L2 penalty are the best choice in generative and discriminative models respectively as they were stable among all experimental classifications. As observed in previous studies [20]- [22], generative classifiers (MB and Markov) are sensitive to how they infer their parameters (class-conditional densities), either by MLE or Bayesian approaches. MLE approach could overfit and produce a sparse parameter matrix W when unseen k-mers in training step will have null estimates [20], [23] or very small probabilities will underflow the numerical precision [14].…”

Integrative workflows for metagenomic analysis

“…Nonetheless this challenge may be addressed computationally, sorting raw sequencing reads taxonomically (Figure 5 ) and phylogenetically (Weisburg et al, 1991 ; Retief, 2000 ; Darling et al, 2014 ) and thus yield conclusive information about the population of the niche, which can be extended subsequently to the assembled contigs and genes. This process is called taxonomic binning (Droge and Mchardy, 2012 ) and there are numerous tools (Mohammed et al, 2011 ; Pati et al, 2011 ; Luo et al, 2014 ; Wang et al, 2014 ) that rely on homology based or composition based approaches (Rosen and Essinger, 2010 ).…”