You say there are few selective sweeps, but what do you compare with? Did humans have less selective sweeps that chimps or average mammal species at same time, or did humans at 55 kya had less selective sweeps than humans at 8 kya?
I think the three loci detected by Field et al. or the twelve loci detected by Palamara et al. are just objectively a small number no matter how you slice it, especially given that most common traits are driven by thousands of variants and there are very many traits. I'm a human geneticists but my lay understanding is that other species often have a much larger number of sweeps as well. Unfortunately we do not have great methods for detecting sweeps older than 50kya and so it is hard to say whether selection has been slowing down or has always been slow.
I don't know anything about statistical genetics so I probably misunderstood something, but I found this paper https://www.pnas.org/doi/pdf/10.1073/pnas.2213061120 also coauthored by Tobler and Souilmi (and others). In Table 1 (page 5), the authors indicate that one-third of the "genes inferred as under positive selection" are associated with neuronal functions. On page 4, the authors write: "Intriguingly, eight of the ten ancient Eurasian neuronal genes are associated with severe intellectual disabilities and developmental delay phenotypes in humans".
Also, in this preprint https://www.biorxiv.org/content/10.1101/2023.02.05.525539v2.full, the authors identified a "young peak of genetic variants arising at ~280,000 until ~2,000 years ago reaching a maximum at ~55,000 years ago". Then, they write that "genetic variants related to ‘Psychiatric’ phenotypes presented an evolutionary age younger than expected by chance (median evolutionary age = 412,639 years old), a class of traits suggested to be linked to human brain evolution."
How is this consistent with your claim that "Most studies agree that instances of locus-specific selection typically influence phenotypes related to immunity, pigment, and diet with little to no instances for cognitive/behavioral phenotypes"? Are these two studies wrong? Or did I just misunderstood everything?
Thanks for the comment. It is definitely the case that you can find papers showing some selection on some brain genes somewhere -- people are interested in this hypothesis and they will keep looking and reporting findings. I think important questions to ask are: (a) is this a large and well powered study that accurately addresses confounding?; (b) how definitively are the variants/genes linked to the mechanism?; (c) are they testing against a proper background and null. I tried to highlight studies that are in reputable journals (or pre-prints from the Reich lab that will likely end up in reputable journals) and satisfy these criteria. Of the studies I looked at the consensus was no clear effects on the brain and no enrichment for brain effects (most clearly demonstrated in the Akbari et al analysis and Reich's discussion of it). I think this is broadly reflective of where the field is in 2024.
To the papers you mentioned. Tobler and Soulimi PNAS do not, in my opinion, have a good answer to questions (b) or (c). The mechanistic connection from sweep to gene is very uncertain (they are basically just looking for genes that are near or overlap the detected sweep regions, which is no guarantee that the sweeping allele is actually operating on that gene) and the connection from genes to neuronal function is also quite speculative: they are looking for any entries in the literature or in rare disease databases that have to do with the brain, which is very easy to find just by chance. They do not do a background or null comparison that I could find to actually evaluate if the co-occurrence with brain genes at these loci are more than expected by chance given their protocol.
With respect to the other pre-print, it is conducting a polygenic analysis over a large number of variants rather than looking at individual loci, and this type of analysis is just very hard to do right without confounding. It's something I plan to write about in the future (a similar analysis is conducted in Akbari et al.). My skepticism also goes up whenever these analyses do not show enrichment on pigment as a positive control (something Akbari et al. *did* show), since that is one class of traits where we know quite clearly that selection has occurred.
You say there are few selective sweeps, but what do you compare with? Did humans have less selective sweeps that chimps or average mammal species at same time, or did humans at 55 kya had less selective sweeps than humans at 8 kya?
I think the three loci detected by Field et al. or the twelve loci detected by Palamara et al. are just objectively a small number no matter how you slice it, especially given that most common traits are driven by thousands of variants and there are very many traits. I'm a human geneticists but my lay understanding is that other species often have a much larger number of sweeps as well. Unfortunately we do not have great methods for detecting sweeps older than 50kya and so it is hard to say whether selection has been slowing down or has always been slow.
Thank you for this interesting post.
I don't know anything about statistical genetics so I probably misunderstood something, but I found this paper https://www.pnas.org/doi/pdf/10.1073/pnas.2213061120 also coauthored by Tobler and Souilmi (and others). In Table 1 (page 5), the authors indicate that one-third of the "genes inferred as under positive selection" are associated with neuronal functions. On page 4, the authors write: "Intriguingly, eight of the ten ancient Eurasian neuronal genes are associated with severe intellectual disabilities and developmental delay phenotypes in humans".
Also, in this preprint https://www.biorxiv.org/content/10.1101/2023.02.05.525539v2.full, the authors identified a "young peak of genetic variants arising at ~280,000 until ~2,000 years ago reaching a maximum at ~55,000 years ago". Then, they write that "genetic variants related to ‘Psychiatric’ phenotypes presented an evolutionary age younger than expected by chance (median evolutionary age = 412,639 years old), a class of traits suggested to be linked to human brain evolution."
How is this consistent with your claim that "Most studies agree that instances of locus-specific selection typically influence phenotypes related to immunity, pigment, and diet with little to no instances for cognitive/behavioral phenotypes"? Are these two studies wrong? Or did I just misunderstood everything?
Thanks for the comment. It is definitely the case that you can find papers showing some selection on some brain genes somewhere -- people are interested in this hypothesis and they will keep looking and reporting findings. I think important questions to ask are: (a) is this a large and well powered study that accurately addresses confounding?; (b) how definitively are the variants/genes linked to the mechanism?; (c) are they testing against a proper background and null. I tried to highlight studies that are in reputable journals (or pre-prints from the Reich lab that will likely end up in reputable journals) and satisfy these criteria. Of the studies I looked at the consensus was no clear effects on the brain and no enrichment for brain effects (most clearly demonstrated in the Akbari et al analysis and Reich's discussion of it). I think this is broadly reflective of where the field is in 2024.
To the papers you mentioned. Tobler and Soulimi PNAS do not, in my opinion, have a good answer to questions (b) or (c). The mechanistic connection from sweep to gene is very uncertain (they are basically just looking for genes that are near or overlap the detected sweep regions, which is no guarantee that the sweeping allele is actually operating on that gene) and the connection from genes to neuronal function is also quite speculative: they are looking for any entries in the literature or in rare disease databases that have to do with the brain, which is very easy to find just by chance. They do not do a background or null comparison that I could find to actually evaluate if the co-occurrence with brain genes at these loci are more than expected by chance given their protocol.
Just for fun, I looked at the first five genes on their list of "neurological" genes in Table S6A: AMBRA1 (https://en.wikipedia.org/wiki/AMBRA1#Clinical_significance) which is related to cancer, autoimmune disease, and Alzheimers; ARFGEF1 (https://en.wikipedia.org/wiki/ARFGEF1) which has no clear function; COL4A3BP (https://en.wikipedia.org/wiki/COL4A3BP#Clinical_significance) which is linked to autoimmune conditions; DOCK3 (https://en.wikipedia.org/wiki/Dedicator_of_cytokinesis_protein_3) which is linked to neurological disorders; and DOK5 (https://en.wikipedia.org/wiki/DOK5) which has no clear function. This is not a rigorous analysis or anything, but it aligns with my prior that the connection to neurological conditions is still very speculative. To be clear, I don't think the authors did anything wrong, they are pretty clear that these genes should be taken with a grain of salt.
With respect to the other pre-print, it is conducting a polygenic analysis over a large number of variants rather than looking at individual loci, and this type of analysis is just very hard to do right without confounding. It's something I plan to write about in the future (a similar analysis is conducted in Akbari et al.). My skepticism also goes up whenever these analyses do not show enrichment on pigment as a positive control (something Akbari et al. *did* show), since that is one class of traits where we know quite clearly that selection has occurred.