I wonder if it could be a good and easy approach even this one: On a dataset of metagenomic data I run both Pathofact than a software for taxonomic classification (such as Kraken2 or Metaphlan3); then, once i get the identification of bacterial toxin for example I can merge the two information and so associate the toxin to the bacteria identified by the taxonomic classifier and so assign to the toxin the rel abundance of the identify bacteria? what do you think about? May it be a good approach? Thanks