I think this emphasizes the need for more concrete models of pleiotropy to help us know when these verbal models should even apply in principle (let alone in expts like @oliviamghosh.bsky.social's). In that vein, was very excited about the new work that @djhelam1.bsky.social‬ talked about at NITMB

Thank you for all your questions! I think they really point at some of the important challenges in interpreting this data.

I think it would be super cool to explore some of these assumptions and models theoretically, but (spoiler alert) we don't actually find evidence for the pleiotropic expansion model in our data! So perhaps these assumptions are not borne out in reality (in our system at least).

For the purposes of our paper, we are using it as a catch all for models in which the evolution condition is "special" from this top-down perspective. Specific mechanisms that would give rise to this are pretty subtle, as we have discussed here!

I think this raises the point that this pleiotropic expansion model is actually a little bit tricky to implement concretely, and definitely requires some additional assumptions.

this might mean that actually to make this model work, you do need some qualitative difference in E1 and E2 from the organism's perspective! ie how many traits are already optimized in each, and how those overlap with each other.

In E2, these yellow traits may not be optimized, so they are likely to be affected by adaptive mutants in E2. Then, when measured in E1, they will show up as additional detectable traits, as long as green traits were also not optimized in E2.

So you can imagine a scenario where there are certain traits that are already optimized in E1 (let's say these are the yellow traits). No adaptive mutant will affect these traits because changing them would decrease fitness (ignoring subtleties around the net fitness effect given multiple traits)

There are some implicit assumptions that must be true for this pleiotropic expansion model to work. First, just to clarify, the fact that a box is colorful does not mean it has a positive effect on fitness, it just means it is relevant. So a mutant's affect on a trait can be good in E1, bad in E2

You are bringing up an interesting (and somewhat subtle) point! I think if I understand your confusion properly you are saying that if there were these yellow traits that mattered to fitness in E1 all along, why didn't the mutants evolved in E1 ever affect them?

We expand on this more in the new text, so feel free to check it out! Also happy to chat more offline if it is still not clear.

So by that argument, if we looked at mutants evolved in E2, we would expect them to affect few traits in E2, and then many traits in E1.

But the "pleiotropic expansion" model says that the home environment is special. In the evolution environment, these mutants are *conditionally* low-dimensional.

On the assumption that mutations affect many traits and are generically pleiotropic, in most environments they will appear so, and hence we would discover a relatively "high" dimensional space by observing their fitness variation around that environment.

But the basic idea of what you said is right – there is no qualitative difference between environments 1 and 2. Instead, the "low-dimensionality" of home mutants is more of a statement about ascertainment bias.

Thanks for taking a look! We actually have an updated v2 manuscript that I think clarifies our two competing hypotheses in figure 1:

www.biorxiv.org/content/10.1...