Hi, I ran magmap 1.2.0 to map two metatranscriptomes to GTDB release 226 and this worked flawlessly. However, when I add samples (currently trying with 12),

BBMAP_ALIGN

requires copious memory. I used > 15,000 core hours on

BBMAP_ALIGN

alone and only 3/12 finished. I used the parameter

sequence_filter

in combination with

SILVA_138.1_allrRNAref_NR99_tax.fasta

to try and reduce memory usage. Are there other things I could try? I am considering to lower the number of metatranscriptomes from 12 to 6 and try again. With 12 meta-t's I am mapping to 18,862 genomes, but this number goes down when lowering the no. of samples.

The number of genomes decide the size of the database you're mapping to. If you have samples that do not overlap much that means you get a very large set of target genomes and a large database since by default all samples are mapped against all genomes. If you set the

--genomeset_mode

to

joint

(see nf-co.re/magmap/1.2.0/parameters#…) mapping will occur against a sample specific set of genomes instead which should reduce memory requirements. The total genomes should remain the same.

That helped a lot. Many thanks

Although there was a (grave) mistake: it should be

sample

of course,

joint

is the default.

Yes, I understood. I will rerun next month 😬