Skip to content

Terminology

Info

Even though we have very specific concepts in mind when talking about the different tools and their results, we don't (yet) apply this terminology consistently everywhere. We will do so as time permits. In the meantime, please study the concepts presented herein carefully and form your own judgement how they apply elsewhere in the documentation.

Our thinking about terminology is strongly influenced by the categorisation proposed by Sun et al. (2021)1. To summarise: Broadly, we call all of these tools metagenomic profilers. They can be more finely grouped by considering their type of reference database and thus the kind of relative abundance profile produced. Furthermore, the underlying method for sequencing read classification impacts the computational complexity (time and database size/memory) dramatically.

Criteria

Type of Abundance/Profiler

Our main criteria for categorizing a metagenomic profiler are summarised in the list below, as well as shown in the schema.

  • Reference database:
    • DNA-to-DNA, aka a metagenome library
    • DNA-to-AA, aka a genomic library of protein coding sequences (amino-acid)
    • DNA-to-marker, aka marker gene, aka gene families selection
  • Result:
    • relative sequence abundance
    • relative protein sequence abundance
    • relative taxonomic abundance
graph TB

    P(Metagenomic Profiler)
    D(DNA-to-DNA)
    A(DNA-to-AA)
    M(DNA-to-marker)
    S(Sequence Abundance)
    SAA(Protein Sequence Abundance)
    T(Taxonomic Abundance)
    TP(Taxonomic Profiler)
    SP(Sequence Profiler aka Classifier)

    P --> D
    P --> A
    P --> M
    D --> S
    A --> SAA
    M --> T
    S --> SP
    SAA --> SP
    T --> TP

Differences in Abundance

A key takeaway from Sun et al. (2021)1 is also that the different kinds of relative abundances cannot easily be interconverted by, for example, normalizing sequence abundance by genome size. Reasons for this include, in particular for metagenome-assembled genomes (MAGs), inaccurate genome size, unknown polyploidy. Sun et al. (2021)1 observed that broadly sequence abundance tends to overestimate large genomes and underestimates smaller genomes. Any attempts to perform normalization by, for example, genome size, are just as likely to introduce systematic errors than improve the results. Furthermore, α-diversity measures appear to be statistically significantly affected by the type of relative abundance - higher for taxonomic abundance. The type also affects β-diversity although the effect is harder to quantify.

Classification Method

Additionally, we may be interested in the method used to classify sequencing reads.

  • Method:
    • index
    • alignment

Categorisation

Tool Reference Database Result Method
Bracken See Kraken2 See Kraken2 Bayesian re-estimation of sequence abundance
Centrifuge DNA-to-DNA Sequence abundance (optionally normalized by genome size) FM index
DIAMOND DNA-to-AA Protein sequence abundance
Kaiju DNA-to-AA Protein sequence abundance
Kraken2 DNA-to-DNA (DNA-to-AA) (Protein) sequence abundance k-mer index
KrakenUniq DNA-to-DNA Sequence abundance see Kraken
MEGAN6/MALT DNA-to-DNA Sequence abundance alignment
MetaPhlAn DNA-to-marker Taxonomic abundance alignment (bowtie2)
mOTUs DNA-to-marker Taxonomic abundance alignment (bowtie2)

  1. Sun, Z., Huang, S., Zhang, M., Zhu, Q., Haiminen, N., Carrieri, A. P., Vázquez-Baeza, Y., Parida, L., Kim, H.-C., Knight, R., & Liu, Y.-Y. (2021). Challenges in benchmarking metagenomic profilers. Nature Methods. https://doi.org/10.1038/s41592-021-01141-3 ↩↩↩