Explanation of species-level classifcation filed "s_unknown_species"

[your-highness]

Dear Temesgen,

I need some clarification on the output of slimm species-level classification "_profile.tsv". For a test sample I get the following output:

taxa_level      taxa_id linage  abundance       read_count
species 9606    k__Eukaryota|p__Chordata|c__Mammalia|o__Primates|f__Hominidae|g__Homo|s__Homo sapiens   89.1426 25602864
species 45219   k__Viruses|p__unknown_phylum|c__unknown_class|o__unknown_order|f__Arenaviridae|g__Mammarenavirus|s__Guanarito mammarenavirus    0.0178544       5128
species 1821749 k__Viruses|p__unknown_phylum|c__unknown_class|o__Picornavirales|f__Picornaviridae|g__Cardiovirus|s__Cardiovirus A       0.0136867       3931
species 0*      k__unknown_superkingdom|p__unknown_phylum|c__unknown_class|o__unknown_order|f__unknown_family|g__unknown_genus|s__unknown_species       10.8259 3109321

While most reads are classified as Human (89.14% of the reads), 11% of the reads are classified as unknown species. This is confusing because these reads are contained in the BAM file and must be mapped to a reference genome (bowtie2 --no-unal).

Does the fraction of 11% correspond to one species? Or: Could the species not be resolved because of missing taxonomic information? Or: Are these reads not discriminative for 1 species?

All the best,
Johannes

[temehi]

Does the fraction of 11% correspond to one species? Could the species not be resolved because of missing taxonomic information? Or: Are these reads not discriminative for 1 species?

It looks like, these reads are not discriminative, and could not be resolved even at (super) kingdom level, which means that they are mapped both to a virus(es) and human genomes. This might happen as some viral DNA fragments are known to exist within the human genome. But 11% is surprising.

Could you try to do the classification at (super) kingdom level and share the results?

Best,
Temesgen

[agakrawczyk]

Hello Temesgen!
Thank you for the answer!
I used the same sample as Johannes.
Here is a report from the run where I set the classification to the superkingdom level
taxa_level taxa_id linage abundance read_count strain 0* k__unknown_superkingdom|p__unknown_phylum|c__unknown_class|o__unknown_order|f__unknown_family|g__unknown_genus|s__unknown_species|r__unknown_strain 100 28721244

and I also tried the phylum level

taxa_level taxa_id linage abundance read_count phylum 7711 k__Eukaryota|p__Chordata 89.1427 25602891 phylum 2759* k__Eukaryota|p__Eukaryota_unclassified 9.7433 2798398 phylum 0* k__unknown_superkingdom|p__unknown_phylum 1.114 319955
What I find weird is that on the kingdom level nothing was classified as we see in the report with species level. Do you have any idea why is it like this?

I used band-width 80, cause our reads have the average length of around 76 bp. (--cov-cut-off not specified, so it was default one)

All the best,
Aga

[temehi]

Sorry guys for not being responsive. Had really a busy month. I am working on it and will update you!

[temehi]

The problem looks like due to the broken lineage from kingdom to species.
k__Viruses|p__unknown_phylum|c__unknown_class|o__Picornavirales|f__Picornaviridae|g__Cardiovirus|s__Cardiovirus A

The whole hierarchical assignment to the LCA depends on having a 7 level taxonomic tree.
our virus Cardiovirus A, for example, has this lineage in the database:

super-kingdom;   no rank;          order;         family;       genus;       species 
      Viruses; Riboviria; Picornavirales; Picornaviridae; Cardiovirus; Cardiovirus A

This breaks the LCA assignment as we do not have any phylum or class in the lineage.

[your-highness]

This is a problem, indeed.

Do this mean that for all species where the lineage is incomplete the LCA computation fails?

[your-highness]

Hi @temehi ,

Does the LCA computation for all species with a broken lineage fail?

Will there be a fix soon?