Right. To play devil’s advocate, I think Sanford’s argument would be that for most large multicellular eukaryotes, they have much smaller effective population sizes compared to bacteria, and much larger genomes, and significantly higher mutation rates, and that these in combination should cause inevitable fitness decline in (for example) humans and so on.
So the MA program is incapable of simulating GE for realistic effective population sizes for bacteria, but that it is capable of doing so for mammals, for example. And here his program does show fitness decline.
So now the question becomes what, if anything, is wrong with how Sanford simulates evolution for populations that are within it’s range of sizes, or whether these really are destined to inevitable fitness decline? I’ve already given numerous arguments in the other threads about what I think he gets wrong (the DFE doesn’t stay constant, for example).
That is not a good argument. We all have much larger effective population sizes than GA can model.
That just isn’t true.
What is interesting is that it seems GE would predict on threatened species, with very low total populations, should face catastrophic failure. However, we have seen that even when those populations are low, they can sometimes still persist, but I don’t have data on this. It seems this would be a good way to look at this.
It is worth emphasize though that fitness is different than virulence or longevity. It doesn’t make sense to equivocate those concepts either…
That makes sense–as it is literally the basis of Neutral Theory…
If we assume the mean deleterious DFE is 0.001 on an exponential distribution, like Paul has mentioned, the probability of seeing a VSD in humans that cannot be selected (s < 0.00005) is 4.9% at Ne = 10,000 and (s < 0.000016) 1.6% at Ne = 30,000.
If we charitably assume every single mutation falls in the deleterious category [a completely absurd and demonstrably false claim], you will get 1.6-4.9 VSDs per 100 mutations.
If at any time, those VSDs cumulatively lower the fitness between 1/20,000 – 1/60,000, they will no longer be able to achieve fixation in the population through genetic drift and will be subject to purifying selection. GE does not explain how it violates this speed limit to generate a strongly deleterious genotype in all members of the population simultaneously causing extinction. GE additionally does not explain why we cannot detect the other > 95.1% (1 - % of VSDs) of purportedly deleterious mutations predicted to exist under this DFE. We should be able to quite easily see those 95 / 100 mutations that are patently deleterious in pedigree studies–yet we don’t.
I would say that quite readily demonstrates the DFE being used by GE is factually incorrect. Although, we already knew the DFE in Kimura’s work was for the deleterious coding spectrum anyway…
It’s not uncommon for published simulations to use unrealistic population parameters (e.g., size), but the whole suite of parameters used in the model must then be properly scaled to match the unrealistic parameter.
Okay so effective population size does seem to make a difference in MA, but the Ne differences appear to have to be rather large before the change in rate of fitness decline becomes noticeable.
No, Ne is not the census size – it’s the effective size, i.e. the size of an ideal population in whatever model you’re using that matches the behavior (whichever behavior you are measuring) of your real population.
I think you’re just flat out wrong this time. This is why ‘2Ne’ so commonly shows up in pop gen formulas for diploid species – that is the number of genome copies.
It would be true if all the deleterious mutations were inherited as a single linkage group that an organism either possesses or doesn’t. And there would have to be organisms of both sorts in the population.
I am willing to believe you, but I would really appreciate @Joe_Felsenstein’s confirmation. It seems that this is not consistent in some of the population genetics software. So by usage, there seem to be multiple conventions. If some of those programs are flat out wrong, and it isn’t a matter of convention, I’d like to know for certain from someone with normative authority in this space.
Correct. I am not claiming Ne = Nc. Ne represents the number of individuals contributing to the heterozygosity of the census population i.e.–the number of individuals producing the same amount of genetic drift as the census population Nc.
There are multiple ways to calculate Ne–and it’s possible there might be a number of interpretations for different models.
We should probably revisit this issue then as I’m not sure the explanations given adequately address the underlying mathematics of what happens in real populations.
Selection can act on the alleles and arguably must act on the alleles–but I’m not sure we are interested in discussing intragenomic conflict / Selfish Gene or the operationalized assumptions (normally some idealized HWE postulate) we make in ecology. That tends to get heated and neither perspective is entirely wrong or right.
Because they are using a convention where Ne is to individuals, which creates an ugly two in the formulas, and makes it so they only apply to diploid populations, not other sorts of populations.
By the way, I have never examined the source code for Mendel’s Accountant, but some years ago I read the documentation. The things to to look at, which struck me a odd back then are (1) the scheme for linkage, which looked likely to produce a strong Muller’s Ratchet effect as back then it involved a lot of chromosomes each of which had no recombination within it, and (2) the procedure for selection, which sounded very weird and not like the way selection is done in Wright/Fisher models in theoretical population genetics.
But that was many years ago and they may have changed the program since.
