Enterotypes-2018

The recent update on enterotypes¹ was an important read.

Yes, we desperately need to reduce the dimensionality of the gut microbiome data and discover the stable “archetypes” of the microbiome functional states.

The concept of genera-based enterotypes is a step in this direction. However, one may feel a “sense of fragility” while reading the 2017 update paper. And it is not only related to inter-study and inter-clustering-method variations of the results but – fundamentally – to the type of the data taken as a base of such classification.

While genus-level classifications look as a valid approximation of functional-level differences on the surface, one can potentially question the very base of this assumption.

Let’s imagine what will happen when such clustering will be eventually done on the basis of the whole-metagenome data in a large enough scale?

We may foresee 2 strategic possibilities:

Hypothesis #1. The fine-grained genomic data will refine the structure of the genera-based enterotypes, while the overall, bigger picture enterotype landscape will stay the same.

Hypothesis #2. The finer-resolution metagenomic data will reveal entirely different (or loosely correlated to the genera-based enterotypes) stable microbiome states, where the “stability core” will be related to deeper functional patterns, transcending the phylogeny level.

Let me introduce a remote analogy. Let’s imagine that some aliens started to study human population on this planet. What “subtypes” would they see right away? – Obviously, skin color. And since the spatial distribution of humans also tends to correlate with the skin color, this would be considered by aliens as “substantial” evidence that the skin color is a primary functional characteristic of the human beings.

And we know that what is really “functional” may lie in a totally different dimension.

This may explain the fragile nature of the current enterotypes: if they, in fact, represent a shadow of real deeper-functionality “archetypes” of community composition via (limited) functionality-genera correlations, then the projection of the real (functional) enterotypes onto the visible (genera-based) enterotypes may be highly non-linear, resulting in falling of the classification apart in a number of cases.

So, intuitively I am inclined to the Hypothesis #2. However, the time will tell.