187 — Optimized Bacteria are Environmental Prediction Engines

Marzen & Crutchfield (1802.03105)

Read on 23 February 2018
#bacteria  #computation  #prediction  #optimization  #biology  #markov-process  #POMDP 

This paper discusses the ability of even simple organisms — such as bacteria — to hedge their evolutionary “bets” when it comes to phenotypic variability in an isogenic (or genetically identical) population.

At first, it seems as though identical genotypes should yield identical behavior in single-celled organisms in the same environment. But if that environment were to change, as real-world conditions often do, then a perfectly uniform species would quickly die out, because it only anticipated one environment.

Instead, the authors show that previous environemnts (dubbed the epigenetic memory) have an impact on the bacterium, and the sensed environment (which may differ between otherwise identical bacteria) can also affect the genetic expression of an individual organism.

The authors demonstrate (mathematically) that the way to optimize for reproduction is for a population to express a noisy prediction of the environment. One interesting byproduct of this conclusion is that it seems as though having a noisy sensor — being unable to perfectly sense the surrounding environemnt — is evolutionarily beneficial for a population.

Maybe that’s profound and we should learn something about society from this?

Or maybe this is just an interesting fact about bacteria.