Oh very cool—I think I missed that paragraph in your plos bio paper! Super interesting that these effects seem to matter in both yeast and E. coli

I view genetic drift and decoupling noise as more fundamental demographic stochastic forces, which go on to affect downstream and emergent dynamics.

I think that is fair to say in one sense. The distinction I want to make is that genetic draft is emergent from an interplay of mutation, selection, etc. Changing population genetic parameters, including the strength of drift or decoupling noise, would also change genetic draft.

Thread from the preprint 👇
bsky.app/profile/joao...

Yeah, genetic drift is dominant source of fluctuations at low frequencies, but then decoupling noise starts to dominate above frequencies ~1/(δ*N_e). So depending on the parameters, that cross-over point can be at a really low frequency. I don't know about recombination though, great question!

If you’ve gotten this far, thanks for reading and we welcome any feedback that you might have!

We spend a lot of time trying to measure fitness effects in evolution experiments, but comparatively little effort measuring the noise. I think that it is time to pay more attention to the fluctuations!

When we think of evolutionarily-important stochasticity, we usually think of genetic drift. But decoupling noise is like the shy cousin of drift—largely overlooked, but an important and likely common source of randomness in the frequencies of closely related genotypes.

Finally, we develop some new popgen theory. Some key findings: (1) Decoupling noise can significantly shift the ability of natural selection to distinguish between different fitness effects (2) Decoupling noise can leave selection-like signatures in the SFS

Because N^2-scaling abundance fluctuations are common across populations, we also think that decoupling noise may be ubiquitous. For example, we also find signatures of decoupling noise in the barcoded yeast experiments from the Petrov and Sherlock lab

The characteristic (Lyapunov) time is pretty fast—about 5-10 hours. So the dynamics look effectively stochastic if we’re taking samples every 24 hours. Only with these densely sampled time courses can we see the chaos.

So what is the cause of these fluctuations?

We cultured replicates and tracked the populations over a 24 hour cycle. The replicates exponentially diverge from each other! This is the signature of chaotic dynamics—small differences between replicates are exponentially amplified

Large frequency fluctuations may not be surprising if we were in a noisy environment. But we’re trying as hard as possible to maintain a constant environment, using closely related genotypes!

This is similar to previous models that invoke a fluctuating environment, but we know that many other mechanisms can cause these types of abundance fluctuations (e.g. chaos, aggregation, etc.)

But f^2-scaling frequency fluctuations don’t arise unless the abundance fluctuations are decoupled (to some degree) between the genotypes in the population. So we call these frequency fluctuations “decoupling noise”.

How do we explain this?

We developed a flexible model that can account for the scaling behaviors. Uncorrelated offspring number fluctuations causes classical genetic drift. In contrast, correlated offspring number fluctuations cause ~N^2-scaling abundance fluctuations.