I have a subset of high-quality MAGs from different environments, sequenced on different runs. I want to know, for any particular MAG, if it (or something similar to it) is present in any particular sample.

I've recently started learning anvi'o for metagenomic visualization. One of its perks is that it can nicely represent coverages for either contigs or bins (tutorial example: https://merenlab.org/images/miscellaneous/2017-05-11-anvi-refine-by-veronika/Image_1.png , the circles marked Sample 07 and 08). Example_plot

The coverage options are defined as follows:

Mean coverage
Average depth of coverage across contig.
Add up the coverage of each nucleotide in a contig, and divide by the length of the contig. This is the default view in most cases.

Mean coverage Q2Q3
Average depth of coverage across contig excluding nucleotide positions with coverage values falling outside of the interquartile range for that contig.
Calculated the same as mean coverage, except only incorporating those nucleotide coverage values that fall within 2nd and 3rd quartiles of the distribution of nucleotide coverages for that contig. This can help smooth out the mean coverage visualization by removing nucleotide coverage values from the equation that may be outliers due to non-specific mapping.

Additional representation options:

Mean coverage of a contig divided by that sample’s overall mean coverage.
Abundance for a contig is constrained to within one sample. In a sense this view is telling you that those contigs with larger abundance values are more represented in that sample (i.e. recruited more reads) than those contigs with smaller abundance values. And it does this by providing the ratio of that contig’s mean coverage to that sample’s overall mean coverage incorporating all contigs. So if your abundance value is 2, that contig’s mean coverage is twice that of the mean for all contigs in that sample.

Relative abundance
Proportion of reads recruited to a contig out of the total reads recruited to that contig across all samples.
Relative abundance considers one contig across all samples. This will tell you in which sample a particular contig recruited the most reads. Since it is normalized to the total number of reads recruited to contig across all samples, values from all samples for a given contig will always sum to 1.

Max-normalized ratio
Proportion of reads recruited to a contig out of the maximum number of reads recruited to that contig in any sample.

This also considers one contig across all samples. But in this case the value is normalized to the single maximum value for that contig (as opposed to the sum as in relative abundance). Because of this the sample containing the contig that contributed the max value will always equal 1, and the value for that contig in the other samples will be the fraction of that max.

I'm trying to cross-map the reads between samples to the best MAGs (+within the same sample). The reason I'm trying to map to MAGs as opposed to genomes, as usual, is that some of my MAGs of interest represent genera (at the least) without available genomes, so it feels like I'd lose considerable information that way. But then, I'm learning this as I go along, so for all I know it might be a bad idea.

From what I understood, these settings generally make sense / are used for samples that have been co-assembled. However, if I've been assembling samples independently and co-binning (due to wanting to avoid chimeras) or I want to compare some of the metagenomic composition of different environments (which really wouldn't make sense to either co-assemble or co-bin), would it still make sense to use coverages as a proxy for presence / absence / abundance?

Thank you for your time!

  • 1
    $\begingroup$ 1. any normalization? 2. have you cross-mapped reads between samples? Basically, for metagenomics, if you want to compare abundance between samples, you usually have a huge genome database that you map different samples' reads to, and then you use that to tell the difference between samples in a more or less unbiased way. Is that what you are using anvi'o to do? $\endgroup$ Jan 16, 2022 at 21:54
  • 1
    $\begingroup$ @MaximilianPress 1) I don't know. Would something like 'rel. Abundance' (added above) count? 4) Yes/no. $\endgroup$
    – Laura
    Jan 17, 2022 at 6:24


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.