An interesting observation is the decline in brain size in recent millennia. This followed a very long rising trend in brain size, suggesting that selection against smaller brains was operative until quite recently.
As we collect more ancient DNA sequences these types of clever tests for natural selection will increase in power. For some reason the race “scientists” believe that actual human geneticists are “afraid” of these advanced modern human genetics studies. Quite to the contrary, the community eagerly anticipates more powerful studies taking advantage of new genome sequences, ancient and modern, as well as more refined and robust methods which obviously excludes classical twin studies whose main advantage was that it was the best we could do in the Dark Ages before we had any genome sequence.
As a teacher I sometimes have a student who doesn't go to lecture and doesn't do the reading. Then he will not even ask a question but instead makes a statement that demonstrates his lack of understanding. Then I will ask him to formulate a question because the very act of formulating a question will force him to acknowledge what he doesn't understand.
Yup, Akbari et al. paper covers a broader time scale and is mostly about polygenic selection so I'm hoping to cover this in a follow-up. I had some initial thoughts on the paper here:
I think it is still very difficult to disentangle how much of what they are seeing is actual selection versus variability from drift/gene flow (as they talk about at the end of the paper).
What’s the internal state of consensus on neutral theory overall? From everything I’ve read it appears like it’s a really good model/default prior. Do most population geneticists feel the same, or is selection dominance still assumed to be the default?
I think the points of consensus are: (a) hard sweeps where new alleles go from zero to fixation in one population are rare to non-existent; (b) background selection, where deleterious alleles are continuously purged from the genome before they have a chance to rise in frequency, is a powerful force; (c) the SDS method has a very elegant way of dealing with population structure and finds very little *very* recent selection. Beyond that I would not say there's much consensus in the field because of uncertainty for how to deal with background selection and population stratification.
What you miss it that neutral selection bias is also a selected for genetic grouping: given some number of genes very often wondered gene will win out by random selection even though it is no better ( but no worse) than all of the others. You need to distinguish between better genes and simply the one which out of many survived randomly.
>How do we make sense of the surprisingly few loci under selection in the past 5,000 years?
Isn't it in the name of your blog? If IQ is determined by 20,000 loci and it correlates at r=1 with fertility, we would gain 6-7 points per generation given current fertility mean and SD, and h^2 = 1/2. Assuming a point in time where all alleles are at p=1/2, IQ = F ~ Bin(n=20,000, p=1/2). W = F/np. E[W|
IQ=10,001] = 1/10,000. But your studies were powered to find s=1/200 or higher, so your studies are underpowered for the infinitesimal.
This post is about locus-specific selection, the word "locus" or "loci" appears 33 times in the text including in the part you cited. I take it you agree then that locus-specific selection is basically negligible in recent history and there are no individual IQ genes with significant genetic differences between populations?
I'm saying most important selection would be too small for your studies to pick up at the locus level. It could still produce massive shifts in the means of polygenic traits like IQ. So from the studies you cited here you can't rule out something like Europeans going up 7 to 15 points in IQ due to selection in the last 1000 years.
Define negligible? I'm saying small but real selection on many loci for highly polygenic traits means selection pressure still changes people a lot in recent history and this is entirely consistent with studies only powered to find s > 0.005 finding almost nothing. New preprint on this posted today: https://www.biorxiv.org/content/10.1101/2024.09.14.613021v1.full.pdf
There are certainly many ways that selection can act on traits that I did not discuss in the article, which focused on recent sweeps. Since each form of selection has unique challenges, I think it's useful to break things down in this way: what's the time scale, how much power does each method have, what false-positives are they susceptible to, what do they find. I'm glad we can agree that there appear to be essentially no meaningful sweeps with s > 0.005 in the past 5,000 years.
suggesting that reported signals of selection after admixture are likely to be false positives because of an insufficient correction for multiple tests., because most geneticists don't know a chi-square from a t-test
heh, maybe, though the issue here was more about incorrectly under-estimating the multiple testing burden genome-wide (i.e. thinking there were fewer independent tests than there really were)
that's what I was thinking, error rate piles up quickly. tbh a pretty interesting study. many years since I thought about this area (long retired) but seem to recall svante pabo (?) and their group finding greater marker (SSRs I believe) similarity among tested individuals belonging to geographic/ethnic distinct groups than within. a result that you just have to retest an md re-retest right? but likely a testament to our ancestors getting around before Delta and Lufthansa 🤣 correct me if I'm recalling that wrong.
An interesting observation is the decline in brain size in recent millennia. This followed a very long rising trend in brain size, suggesting that selection against smaller brains was operative until quite recently.
https://www.frontiersin.org/journals/ecology-and-evolution/articles/10.3389/fevo.2021.742639/full?utm_source=fweb&utm_medium=nblog&utm_campaign=ba-sci-fevo-when-and-why-did-human-brains-decrease-in-size-a-new-change-point-analysis-and-insights-from-brain-evolution-in-ants
As we collect more ancient DNA sequences these types of clever tests for natural selection will increase in power. For some reason the race “scientists” believe that actual human geneticists are “afraid” of these advanced modern human genetics studies. Quite to the contrary, the community eagerly anticipates more powerful studies taking advantage of new genome sequences, ancient and modern, as well as more refined and robust methods which obviously excludes classical twin studies whose main advantage was that it was the best we could do in the Dark Ages before we had any genome sequence.
As a teacher I sometimes have a student who doesn't go to lecture and doesn't do the reading. Then he will not even ask a question but instead makes a statement that demonstrates his lack of understanding. Then I will ask him to formulate a question because the very act of formulating a question will force him to acknowledge what he doesn't understand.
Because of the timing of this post, I thought it was going to be a response to the recent Reich paper, though it was published a day later.
https://www.biorxiv.org/content/10.1101/2024.09.14.613021v1
Are you going to write about this?
Yup, Akbari et al. paper covers a broader time scale and is mostly about polygenic selection so I'm hoping to cover this in a follow-up. I had some initial thoughts on the paper here:
https://x.com/SashaGusevPosts/status/1835685607361896632
I think it is still very difficult to disentangle how much of what they are seeing is actual selection versus variability from drift/gene flow (as they talk about at the end of the paper).
What’s the internal state of consensus on neutral theory overall? From everything I’ve read it appears like it’s a really good model/default prior. Do most population geneticists feel the same, or is selection dominance still assumed to be the default?
I think the points of consensus are: (a) hard sweeps where new alleles go from zero to fixation in one population are rare to non-existent; (b) background selection, where deleterious alleles are continuously purged from the genome before they have a chance to rise in frequency, is a powerful force; (c) the SDS method has a very elegant way of dealing with population structure and finds very little *very* recent selection. Beyond that I would not say there's much consensus in the field because of uncertainty for how to deal with background selection and population stratification.
What you miss it that neutral selection bias is also a selected for genetic grouping: given some number of genes very often wondered gene will win out by random selection even though it is no better ( but no worse) than all of the others. You need to distinguish between better genes and simply the one which out of many survived randomly.
>How do we make sense of the surprisingly few loci under selection in the past 5,000 years?
Isn't it in the name of your blog? If IQ is determined by 20,000 loci and it correlates at r=1 with fertility, we would gain 6-7 points per generation given current fertility mean and SD, and h^2 = 1/2. Assuming a point in time where all alleles are at p=1/2, IQ = F ~ Bin(n=20,000, p=1/2). W = F/np. E[W|
IQ=10,001] = 1/10,000. But your studies were powered to find s=1/200 or higher, so your studies are underpowered for the infinitesimal.
This post is about locus-specific selection, the word "locus" or "loci" appears 33 times in the text including in the part you cited. I take it you agree then that locus-specific selection is basically negligible in recent history and there are no individual IQ genes with significant genetic differences between populations?
I'm saying most important selection would be too small for your studies to pick up at the locus level. It could still produce massive shifts in the means of polygenic traits like IQ. So from the studies you cited here you can't rule out something like Europeans going up 7 to 15 points in IQ due to selection in the last 1000 years.
Are you saying that you do agree that locus-specific selection -- the topic of this post -- has been basically negligible in recent history, or no?
Define negligible? I'm saying small but real selection on many loci for highly polygenic traits means selection pressure still changes people a lot in recent history and this is entirely consistent with studies only powered to find s > 0.005 finding almost nothing. New preprint on this posted today: https://www.biorxiv.org/content/10.1101/2024.09.14.613021v1.full.pdf
There are certainly many ways that selection can act on traits that I did not discuss in the article, which focused on recent sweeps. Since each form of selection has unique challenges, I think it's useful to break things down in this way: what's the time scale, how much power does each method have, what false-positives are they susceptible to, what do they find. I'm glad we can agree that there appear to be essentially no meaningful sweeps with s > 0.005 in the past 5,000 years.
suggesting that reported signals of selection after admixture are likely to be false positives because of an insufficient correction for multiple tests., because most geneticists don't know a chi-square from a t-test
heh, maybe, though the issue here was more about incorrectly under-estimating the multiple testing burden genome-wide (i.e. thinking there were fewer independent tests than there really were)
that's what I was thinking, error rate piles up quickly. tbh a pretty interesting study. many years since I thought about this area (long retired) but seem to recall svante pabo (?) and their group finding greater marker (SSRs I believe) similarity among tested individuals belonging to geographic/ethnic distinct groups than within. a result that you just have to retest an md re-retest right? but likely a testament to our ancestors getting around before Delta and Lufthansa 🤣 correct me if I'm recalling that wrong.
Yeah, in ancient DNA there's extensive evidence of admixture and migration, sometimes shockingly fast.