I guess I am one of the few. We for sure need effective parameters in pop gen. But the idea that there is one scalar that will capture the complexity of population genetics and then can be used for the species as a whole and for every question is bonkers. And has been really damaging imo. - ThreadSky

petrovadmitri.bsky.social • 163 days ago

I guess I am one of the few. We for sure need effective parameters in pop gen. But the idea that there is one scalar that will capture the complexity of population genetics and then can be used for the species as a whole and for every question is bonkers. And has been really damaging imo.

Reposted from Vincent J Lynch 🐘🦣🦥 🦇🐋🐢🐍

What is Ne? Everything.
What has it been hitherto? Nothing. What does it desire to be? Something 🧪 🌱🐋

pubmed.ncbi.nlm.nih.gov/35532202/

Comments

martykardos.bsky.social•161 days ago

Is there a better set of parameters to represent effects of drift?

vsbuffalo.bsky.social•161 days ago

I think the main problem is that Ne doesn't represent drift, but drift and linked selection.

martykardos.bsky.social•161 days ago

Is it not true that selection affects births and deaths and thus the strength of genetic drift across the whole genome (not just selection at linked loci)? e.g. a selective die off is essentially a population bottleneck regardless of the genomic location of the loci encoding a selected phenotype.

martykardos.bsky.social•161 days ago

I never really bought the arguments that effective size varies across the genome due to varying linked selection and recombination effects. does the localized reduction in pi around selected loci not occur IN SPITE of drift, rather than drift being stronger in some regions than in others?

vsbuffalo.bsky.social•161 days ago

My view is that it's useful to imagine a conceptual (not real) "drift-effective population" which determines how fast a neutral allele would experience drift if there were no heritable fitness variation at all in the population, in an idealized WF population (hence why this isn't real). 1/

vsbuffalo.bsky.social•161 days ago

Of course there is heritable fitness variation in every population, and that creates linked selection at different scales, including across the whole genome, which can often be indistinguishable in practice from this conceptual neutral drift. 2/

petrovadmitri.bsky.social•161 days ago

No it’s not true. The same reduction in neutral variation can have completely different effects on weakly deleterious alleles depending on whether it’s driven by 1) bottleneck or 2) BS or 3) draft generated by weak or 4) strong selection. See eg Fig 1 in https://static1.squarespace.com/static/64dd8c893c24b555de0349fd/t/6562f4a143ba84039829422a/1700983971850/88.pdf

vsbuffalo.bsky.social•161 days ago

to clarify: here, I mean the Ne that’s *inferred* from data.

martykardos.bsky.social•161 days ago

Your original statement make more sense with this caveat

vsbuffalo.bsky.social•161 days ago

Hah, sorry. I shouldn’t post while sick with a cold, my thinking is foggy.

petrovadmitri.bsky.social•161 days ago

There is no one “Ne” that represents both

vsbuffalo.bsky.social•161 days ago

I'm not saying there is. I am saying that models often try to estimate an Ne, and when they do, it is determined by both the purely demographic processes (that one could think about as some conceptual drift effective population size) and by linked selection across the genome.

vsbuffalo.bsky.social•161 days ago

A separate issue is that selection perturbs genealogies in ways that cannot be captured by a rescaled Ne in an idealized WF population (I think this is your point?). Either way, Ne doesn't represent anything except in the context of a particular model, IMO.

castellanoed.bsky.social•161 days ago

If all you have is a hammer (Ne), everything looks like a nail (drift).

devoevomed.bsky.social•162 days ago

In grad school I took an evolutionary theory class with Sean Rice, who was an amazing teacher, the most complex #PopGen maths all made sense, but Ne never did, conceptually sure, but in practice it seemed too problematic…

alanrogers.bsky.social•160 days ago

Ne isn't the only parameter in #PopGen, but it will always be important, because it's useful to measure time in units of 2Ne--defined to equal mean separation time btw random gene copies. There will always be such a mean no matter how complex the model.

jmergeay.bsky.social•162 days ago

In the below paper I tackle this issue, and show that the census size is the population size richness, and the effective size the pop size diversity. And that you need both to understand evolutionary responses. https://onlinelibrary.wiley.com/doi/10.1111/eva.70029

petrovadmitri.bsky.social•162 days ago

This helps but I don’t think it can be enough either

jmergeay.bsky.social•162 days ago

I don’t claim that either. Nc and Ne are complementary summary statistics of population size. No more, no less. They are important but not omnipotent.

petrovadmitri.bsky.social•161 days ago

What do you claim?

keholsinger.bsky.social•161 days ago

There isn’t even one Ne. There’s an inbreeding Ne and a variance Ne. Those are (or should be) widely known. Crow and Kimura discussed them extensively. There is also a lesser know eigenvalue effective size. None of them fully capture the dynamics of Wright-Fisher drift.

jmergeay.bsky.social•160 days ago

There are indeed many aspects to Ne, with spatial and temporal components, inbreeding, variance, eigenvalue, coalescence, etc. Which all describe different properties of populations. And yet all are confusingly mixed up in the literature as « Ne ». Out since today: http://dx.doi.org/10.1111/eva.70031

jmergeay.bsky.social•160 days ago

I wrote a paper about it to avoid having to explain this in short skeets and then be blamed for not being complete / exhaustive 😬. And no the paper is complete nor exhaustive. It’s a perspective intended to trigger constructive debate 🤓

Comments

Posting Rules

Reply