Gut microbiome composition does not matter...

...unless you go to the strain level! Or, put differently, what matters is "functional identity", not cladistic assignment. I have always been skeptical of microbiome relative abundance analysis (in the gut or other locations, or even ecological microbiomes) simply because the level of description afforded by current tools, including multi-omics, is not conducive to any firm conclusions: the characterization of a complex ecosystem at the phylum, class, order, family or genus levels, or in the best cases at the species level, can sometimes be correlated with host phenotypes but leads to extreme intra- and inter-individual variability (at that level of description), high uncertainty and hand-waving "explanations" of possible mechanisms.

I have witnessed firsthand the significant differences that may exist between two strains of the same species, for example one that secretes a bacteriocin while the other doesn't, thereby affecting the ecosystem, reducing our ability to understand variability if we only look at the species level -they are the same species.

Instead of "Who is there?", the question we really want to ask is "What are they doing?", which itself may depend on surrounding conditions (host, environmental [pH, T, nutrients] or microbial). But functional analysis, by way of multi-omics and metabolic modeling, is still in its nascent state and can be prohibitively expensive. Strain-level analysis is probably the one method within reach, even though it remains hard and expensive.

So it is with great pleasure that I read a Cell by Cell Press article from a team led by senior authors Nicola Segata and Jingyuan Fu: "Global genetic diversity of human gut microbiome species is related to geographic location and host health." They performed a strain-level analysis for 583 species from 32,152 samples, recognizing that "the substantial intraspecies genetic variability of gut microbes has not yet been comprehensively considered, limiting the potential of linking such genetic traits with host conditions." Even though they did find certain higher-level descriptions to correlate with certain conditions, they "identified 484 microbial-strain-level associations with 241 host phenotypes, encompassing human anthropometric factors, biochemical measurements, diseases, and lifestyle." Interestingly, a lot of these associations are related to age.

I hope that strain-level analysis at scale will soon be the norm, assuming that cost-effective technologies become available. But we have to remember that even with such tools, a lot of "microbial dark matter" still goes undetected, un-isolated and uncharacterized: we have a very incomplete picture of "Who is there", so "What are they doing?" is still the aspiration.