MetaPhlAn 4.2.2 with Jan25 db and generation of the rarefied taxonomic profile

[fasnicar]

The command below is an example for the subsampling (to 25M reads, paired-end) profiling with the new MetaPhlAn 4.2.2.
Notice that with this version, some parameter names changed, as well as the default behaviour now to do unclassified estimation (which can be turned off for generating the profile without unclassified). So, we should update the NF pipeline accordingly.

For the subsampling, we need to do a new mapping of the reads, so that in this way MetaPhlAn can generate the subsamples fastq files.
We can also use these fastq files to produce a subsample of HUMAnN profiles.

metaphlan -1 ${fwd} -2 ${rev} \
    --nproc 4 \
    --input_type fastq \
    --db_dir ${mdb_path} \
    --index ${mdb_version} \
    --force \
    --subsampling_paired 25000000 \
    --subsampling_seed 42 \
    --subsampling_output ${path_sub}/${s} \
    --mapout ${ps}/${s}.bowtie2.bz2 \
    --samout ${ps}/${s}.sam.bz2 \  # CHECK THE LONG VERSION
    --output_file ${ps}/${s}_profile.tsv;

# Rename and compress subsampled reads
mv ${path_sub}/${s}.R1 ${path_sub}/${s}_R1.fastq;
mv ${path_sub}/${s}.R2 ${path_sub}/${s}_R2.fastq;
bzip2 ${path_sub}/${s}_R*.fastq;

Reference: https://github.com/biobakery/MetaPhlAn/releases/tag/4.2.2

[seandavi]

Thanks, @fasnicar. So, are we thinking of performing two sets of analyses for all samples, one with subsampling and one with the complete dataset?

[seandavi]

I'll prepare a distribution of read counts across all samples and we can discuss on a cutoff.