Continuing the discussion from Rarefaction in mothursop:
What about submitting the rarefied results to external software like PICRUSt2? Is this preferential to filtering your OTUs to >= <X_relative_abundance_threshold_perSample>?
i can’t, in good conscience, suggest anyone ever use picrust. it’s conjecture built on conjecture.
What’re the alternatives? Proteomics combined with predictive metagenomics to verify relative functional abundances? Is targeted meta-genomics just too limited in it’s specificity?
Half of my Master’s thesis was built around PICRUSt2, despite knowing its limitations. I do have that icky feeling that makes me want to take a bath, but that was the only solution I could find to “answer” a question with the available data. Any suggestions on how to do metagenomics/metaproteomics well?
The alternative is to sequence metagenomes. Even there it is not a panacea. A metagenome (from sequencing or picrust) will only tell you what we know is there - most genes have no known function. People like to use metagenomes to say what genes are important. But if there are 4000 genes in a genome and that organism goes up, all 4000 of its genes go up, not just one. People pick the gene that helps them tell a story, which is garbage. Metagenomics is useful for understanding functional potential and genetic diversity. I would lean towards transcriptomics because that will tell you which genes are being expressed. Often, the most abundant transcripts come from rare organisms. I think proteomics moves in the right direction too, but the limit of detection is pretty horrible.
Excellent summary! Thanks for providing your perspective.
From what I understand the SomaScan platform offered by SomaLogic/SBI (soon to be Illumina) has a fairly low detection range, but is focused on the human proteome. How do you think SELEX could be used to generate a similar aptamer library that could be used in targeted metagenomic studies?
Just speculating here…