Fig. 1: Workflow.

Taxonomic profiling workflow using Taxometer. After assembling the metagenomic reads, contigs are annotated with any taxonomic classifier. Contigs are processed via Taxometer to extract abundances and tetra-nucleotide frequencies. Taxometer is then trained on the extracted features to predict the taxonomic annotations. A hierarchical loss allows training the model on the annotations from all levels. The predictions are then made for all taxonomic levels for each contig (e.g. completing the annotation from Actinomycetales order to Bifidobacteruim animalis species). For each label, a score is provided.