Can the theory of natural selection distinguish between two correlated phenotypic traits?

Question

Consider the following two situations:

1. A species X has trait $T_1$ that is gradually selected for by some characteristic $E_1$ in some ecological niche but it turns out that the genes for $T_1$ also "turn on" genes for trait $T_2$ so that eventually species X with trait $T_1$ also have $T_2$. Then it turns out that $T_2$ is evolutionarily advantageous i.e it does help in increasing fitness because of some characteristic $E_2$ in that same ecological niche.

2. A species X has $T_1$ that is selected for by some characteristic $E_1$ in some ecological niche, this ecological niche also has characteristic $E_2$ that separately selects for $T_2$.

Both of these cases have as the end product $T_1$ and $T_2$ being advantageous for survival but in one case $T_2$ was selected for and in the other it's usefulness was accidental. What in the theory of natural selection distinguishes these two cases? To put things more bluntly, you are an evolutionary biologist and you are telling the standard story that corresponds to situation 2, how do you know (from the theory of natural selection) that we can discount situation 1?

I did a research project for my undergraduate research where I simulated mutations of cells in a petri dish and I know that mathematically these two cases can be disentangled because they lead to two different cluster size distributions with different scaling laws. As a physicist I know what to do but does a evolutionary biologist using theory of natural selection know what to do.

Update: The point is that in case 1 $E_1$ and $E_2$ could act at different times so that only $T_1$ is responsible for survival and $E_2$ while acting $T_2$ was completely useless for it's survival i.e other species disappeared not because they didn't have $T_2$ but because they didn't have $T_1$. Then a naive biologist could tell separate evolutionary stories assuming that $T_1$ and $T_2$ were selected for separately because when the biologist is observing the situation both $E_1$ and $E_2$ are present and $T_1$ and $T_2$ are present. In other words, he or she would assume case 2 occured. In this case, random mutations and natural selection as told in science classes is completely useless because of $\textit{dynamics}$ in the environment and $\textit{correlations}$ that prevent mutations from being just random.

Why does this bother me? Well, because I hear evolutionary accounts that span hundreds of millions of years and it would seem simply applying the Darwinian story naively could make wrong predictions all over the place. I haven't seen people worry about this as they give evolutionary accounts for everything. From sexual behaviors to marketing.

Answer 1

Are biologists aware of the causation vs correlation problem or are they noobs in statistics?

Evolutionary biologists tend to be pretty good statisticians. For historical consideration, Pearson is the father of the concept of correlation and was a biostatistician and Fisher is (one of) the father of evolutionary genetics and is also a famous statistician (invented the Fisher's exact test, t.test and Anova). Many of the modern developments in modern statistics were brought about by geneticists (incl. developments in HMM, ABC, FDR, etc...).

So, yes we are aware of the causation vs correlation problem.

In the theory of natural selection, can the two situations be disentangled especially when one is talking about evolutionary accounts that are on the order of hundreds of millions of years[?]

Just like in any other causation vs correlation problem, no it is not easy to disentangle the two. If the specific hypothesis put under testing is subject to experimental manipulation, then yes, it is often feasible to disentangle the two. Otherwise, in a purely observational study, one can logically reason and argue for one rather than another causality relationship but in essence, you cannot know in a purely observational study whether a correlation is representing a direct causation.

The follow-up question might then be

Is experimental manipulation feasible in evolutionary biology?

The answer is yes. Not for all questions though.

First there are evolution experiments. An evolution experiment is just like what is sounds like. You put a population in a controlled environment and let it evolve. Repeat the process many time to have a decent sample size. There are also many question of interest to evolutionary biology that can be studied through experimental manipulation (and not only observatory studies) that do not require an evolution experiment.

You seem to be particularly interested in the reconstruction of past evolutionary history of a lineage (evolution biology is much more than a 'historic' reconstruction of life on earth), I would like to highlight that in purely observation studies and in experimental studies not involving evolution experiments, there are plenty of methods to reconstruct past events. Consider for example the loss of genetic diversity at linked variants (selective sweep) caused by positive selection. If two traits are correlated through time (which is very hard to figure out in absence of good fossil records) or correlated over several species (consider having a look at phylogenetic contrasts methods and more modern version of inference of selection over phylogenies), then if selection occurred in a single of the two traits, you should expect to see a differential loss of heterozygosity among the genetic markers explaining ariance for one trait and for the other.

Answer 2

Unless the selective pressure applied by $E_1$ and $E_2$ are applied at different times, it seems like the two scenarios you described are equivalent. In other words, if in both ecological niches, $E_1$ and $E_2$ were present from the start, and both $T_1$ and $T_2$ were present from the start, then absent other factors, the process and result are the same.

However, if you are saying that in the first scenario, $E_2$ was applied only after $E_1$ was selected for, then it is likely that $T_2$ was a genetic hitchhiker, i.e. it didn't provide any advantage before, but the niche has changed and now it provides an advantage. If this is what you're talking about, then you want to look into the concept of genetic drift. Basically, genetic drift is the random mutation of alleles. Genetic drift can lead to speciation if groups of the same species are separated by some sort of barrier to reproduction.

One way to determine if a gene is being actively selected for is to calculate the $K_a/K_s$ ratio which compares the rates of non-synonymous and synonymous substitutions, respectively.

Answer 3

Firstly, I think that you should take a look at the genetic mechanisms behind convergent evolution.

In addition to the above statements, I would say that for your null hypothesis, you would need a situation where if there is a T1 then there is an E1 such that T1 positively regulates T2. Therefore, we can infer that T1 also has T2. I also believe that you can frame this statement using universal instantiation, if you are familiar with either logic or discrete mathematics.

For the second situation in your original question, we could say that there exists a T1 such that characteristic E1 and E2 could result in T2.

I prefer to use the language of logic such that we can build a mathematical model (and perhaps an estimate of all of the different combinations) by integrating the Ka/Ks ratio and the use of some sort of Bayesian analysis or graph theory representation to note the spread of genetic drift and an attempt to delineate the different effects that these mutations could have on phenotypes

Answer 4

First selection does not occur at the species level. Scientist don't talk about a species going extinct due to the lack of a gene.

Second you seem to be confusing what evolution can "tell" with what scientists can tell. The spread of a gene is a statistical correlation occurrence, evolution cannot tell what gene causes a phenotype, genomes are selected for whole hog so to speak, (there are some rare exceptions). It is only through recombination that gene can be segregated, in your example it could be determined that T2 is only active when combined with T1 such situations are well known. indeed this is one of the hypothesis as to why various forms of sex/transfer evolved in the first place, to allow the decoupling and loss of disadvantageous genes.

Third entangled linkage and selective pressures are quite common many traits have multiple effects including counter effects and effects contingent on other genes. What you are describing is a regulatory or linked genes vs independent genes. This is the norm for genetics and is extensively studied and is the reason genetics is not as easy as laymen think it is.Very few genes have only a single selective effect and control genes are quite common.

There is even investigation in the fact that un/successful genes are physically linked by and can effect the survival of other genes in the population just by being physically close to anther gene on the genome and thus likely to be passed on together even with recombination. Entire branches of genetics exist to study linkage.

In your own example there is no accidental usefulness, one is simply a linked or regulatory trait while in the other both are independent. So to answer your question yes evolutionary biologist are quite aware of the issue.

What is told in a classroom, especially a basic classroom is a simplified more abstract version of what is known, in your own field nobody is going to start teaching physics with quantum dynamics, or the relativistic effects on a normal trains mass, you start with simpler and somewhat outdated/incomplete equations until the students can grasp the basic concepts. You are not going to talk about Neutron diffraction with students before they understand what friction and mass are. In a description of the evolution of the avian respiratory system or the function of the human facial recognition system, linkage of individual genes is irrelevant, what is being discussed the dis/advantage or function of the phenotype NOT the genotype that brings them about.