Genome length normalisation + kraken2 confidence score threshold

[fconstancias]

Dear @SilasK,

Thanks for hard work on the field!

I am analysing mouse metagenome dataset and trying different tools & appraoches for an exhaustive characterisation of the taxonomic composition of the fecal metagenomes (i.e., metaphlan4, mOTUS, kraken-braken on CMMG)

I have 3 questions (so far) regarding the kraken pipeline:

• Do you have any recommandation regarding the confidence score threshold --confidence when working with CMMG and mice fecal metagenomes?
• How can I access the length of the genomes you used to build the kraken-braken CMMG databases? Is it possible to retreive the info from the kraken2 database files? I would like to perform genome length normalisation.
• How could I get the taxonomic path of the reference genomes? - to agglomerate data at higher taxonomic ranks.

Thanks for your help.

Best,

Florentin

[SilasK]

Sorry for the delay.

• For the length information you can get them from here: https://ezmeta.unige.ch/CMMG/Supplementary_Tables_and_Figures/Tables/Table_S3_genome_info.tsv

• I just uploaded my begin of making a package with all the utility function for this analysis to CMMG_tools
  There you have the function load_taxonomy

• In order to aggregate the Kraken table you need first make the back mapping from species to representative genome. The following function should do this:

import pandas as pd


def rename_kraken_table(kraken_table,Tax):
    """
        Kraken table that is at species level map to species indexes
    """


    assert kraken_table.columns.is_unique
    
    # create mapping of species to index
    species2ref= pd.Series(Tax.index,Tax.species)
    # remove redundant species
    species2ref = species2ref.loc[~species2ref.index.duplicated(keep='first')]

    kraken_table.columns = species2ref.loc[kraken_table.columns]


```

(Let me test this function once again)

- What exactly do you mean by the confidence threshold? This is a parameter for?

I apologize, I would make the code much smoother but for now I don't have so much time. 

[fconstancias]

Dear @SilasK,

Thanks for the details.

• I will use directly the taxonomical mapping in R using https://ezmeta.unige.ch/CMMG/Supplementary_Tables_and_Figures/Tables/Table_S4_curated_taxonomy.tsv

What exactly do you mean by the confidence threshold? This is a parameter for?

Please see --confidence from the manual or here. It might not be that important for habitat specific database but I was wondering whether you experiemented a bit using cmmg.

Best.

[SilasK]

No I haven't used the confidence score. do you usualy use it?

[fconstancias]

I have been previously working with environmental samples using kraken refseq database and used --confidence of 0.1 based on some results here. Using cmmg on mice GIT metagenomes it might not be needed, I have run the kraken2+bracken pipeline using --confidence values of 0, 0.05, 0.1 and 0.2, and can share the results when I am there.

[jorondo1]

Hi @SilasK and @fconstancias,

I'd love to have your input on this DerrickWood/kraken2#265 (comment)

cheers

[fconstancias]

HI @jorondo1,

Sorry for the delay, check AGalanis97 answer there.