I don’t think that this shows anything different from earlier studies, which it shouldn’t because it doesn’t use different methods.* I guess it simultaneously measures that 25% of the variance is due to common variants and another 25% is due to rare variants in LD with common variants, while previously those results came from separate studies.
I have only skimmed the paper, but I reject the claim in the abstract that these “rare” variants are due to mutation-selection balance (aka mutational load). That is the claim you are talking about, right? They are rarer than the SNPs on the chip, but I think they are too common to be purely deleterious. I don’t see how they could measure such rare variants without sequencing at least some of the subjects.
* One method that is different is that it doesn’t throw out close relatives. This is how it distinguishes common variants from those merely in LD with them. Potentially this could detect the effect of variants rarer than in the earlier article, but it did not.
Having a choice between targeting common variants and rare variants can only make things easier than not having a choice, but why do you think rare variants are easier?
There are tradeoffs between several difficulties. I see three axes. (1) Knowing what variants to target; (2) Cost of edit: (2a) cost of CRISPR and (2b) deleterious effects of the edit; (3) Benefit per gene.
Mutational load likely eliminates (2b). Proofreading the genome allows us to skip (1), but at the cost of doing a tremendous number of edits (2a). Do you really expect that to be easy soon? Alternately, we could try to identify which super-rare variants affect intelligence, paying cost (1), but I think this will be extremely difficult. I expect the effect size (3) to be smaller for rarer variants, although I’m not sure I have a good reason for this.
I have only skimmed the paper, but I reject the claim in the abstract that these “rare” variants are due to mutation-selection balance (aka mutational load). That is the claim you are talking about, right?
Yes
Having a choice between targeting common variants and rare variants can only make things easier than not having a choice, but why do you think rare variants are easier?
Eliminating rare variants almost certainly won’t have negative side effects. If X and Y reduce my intelligence and X is a rare variant and Y a common one, then there is a much bigger chance that Y does something good for me than that X does.
Do you really expect that to be easy soon?
If we have solid evidence that CRISPR editing could result in super-geniuses I expect lots of resources (perhaps in China) to be devoted to the relevant practical problems.
Yes, super-rare variants have some advantages, but as I said, common variants have other advantages. It sounds to me that you have fixated on one approach, rather than considering tradeoffs. For that matter, consider the option of cloning, a mere engineering problem. It won’t get you “super-geniuses,” but that is an arbitrary threshold. It seems to me that you have imposed multiple arbitrary thresholds to isolate this one path.
I don’t think that this shows anything different from earlier studies, which it shouldn’t because it doesn’t use different methods.* I guess it simultaneously measures that 25% of the variance is due to common variants and another 25% is due to rare variants in LD with common variants, while previously those results came from separate studies.
I have only skimmed the paper, but I reject the claim in the abstract that these “rare” variants are due to mutation-selection balance (aka mutational load). That is the claim you are talking about, right? They are rarer than the SNPs on the chip, but I think they are too common to be purely deleterious. I don’t see how they could measure such rare variants without sequencing at least some of the subjects.
* One method that is different is that it doesn’t throw out close relatives. This is how it distinguishes common variants from those merely in LD with them. Potentially this could detect the effect of variants rarer than in the earlier article, but it did not.
Having a choice between targeting common variants and rare variants can only make things easier than not having a choice, but why do you think rare variants are easier?
There are tradeoffs between several difficulties. I see three axes. (1) Knowing what variants to target; (2) Cost of edit: (2a) cost of CRISPR and (2b) deleterious effects of the edit; (3) Benefit per gene.
Mutational load likely eliminates (2b). Proofreading the genome allows us to skip (1), but at the cost of doing a tremendous number of edits (2a). Do you really expect that to be easy soon? Alternately, we could try to identify which super-rare variants affect intelligence, paying cost (1), but I think this will be extremely difficult. I expect the effect size (3) to be smaller for rarer variants, although I’m not sure I have a good reason for this.
Yes
Eliminating rare variants almost certainly won’t have negative side effects. If X and Y reduce my intelligence and X is a rare variant and Y a common one, then there is a much bigger chance that Y does something good for me than that X does.
If we have solid evidence that CRISPR editing could result in super-geniuses I expect lots of resources (perhaps in China) to be devoted to the relevant practical problems.
Yes, super-rare variants have some advantages, but as I said, common variants have other advantages. It sounds to me that you have fixated on one approach, rather than considering tradeoffs. For that matter, consider the option of cloning, a mere engineering problem. It won’t get you “super-geniuses,” but that is an arbitrary threshold. It seems to me that you have imposed multiple arbitrary thresholds to isolate this one path.