I think you should take seriously that in the first paper linked in my comment, the population-wide SNP heritability for cognitive ability is estimated at 0.24 and the within-sibship heritability at 0.14. This is very far from the 0.7 estimate from twin studies. While a perfect estimate of direct additive heritability would be higher than 0.14, I don't think that rare variants (and gene-gene interactions, but this would no longer be additive heritability) would get you anywhere close to 0.7. Note also that UK Biobank with its purportedly poor IQ test represents only ~30% of the sample size in that paper.

Instead, I think it is becoming clear that traditional twin studies made overly strong assumptions about shared and non-shared environments, such that they over-estimated the contribution of genetics to all kinds of traits from height to blood creatinine concentration (compare gold-standard RDR estimates vs twin estimates here). As implied in my original comment, this is likely especially true for traits strongly mediated by society and behaviour. I find it somewhat counter-intuitive, but this kind of finding keeps cropping up again and again in papers that estimate direct heritability with the most current methods.

I don't think that recent ancient DNA papers are affected by this issue, at least not to the same extent. Every aDNA researcher I know is extremely aware of the many pitfalls associated with sequencing ancient material and the various chemical and computational methods to mitigate them. Checking for signs of systematic artifacts in your aDNA data is very routine and not especially difficult.

To provide some brief speculation, I think a major explanation for this paper's errors is that aDNA lab that did the sequencing was quite old, under-staffed, and did not have much recent experience with sequencing human nuclear aDNA, so they had not kept fully abreast of the enormous methodological improvements in this area over the past twenty years.

Basically any paper trying to detect signals of natural selection in humans will turn up something plausibly immune-related [1] [2] [3] [4]. Alongside pigmentation and diet-related genes, it's one of the most robustly detected categories of monogenic selection signal.

While it seems extremely likely that some selection due to pathogenic disease has occurred in humans, I don't think I've seen a paper that convincingly ties a particular selected gene to a particular historical pathogen or pandemic. It would be pretty hard to do so. There's a many-to-many mapping between immune system genes and pathogenic diseases, and selection generally takes many centuries to detectably alter allele frequencies, during which time there have generally been many epidemics and other changes to the environment – which is responsible for the selection signal?

Regarding autoimmunity in particular: I am in no way an immunologist and have much less insight to offer here, but perhaps it isn't surprising (almost tautological?) that immune system genes are often implicated in autoimmunity as well. And I'm not sure that inborn immunity due to HLA alleles or similar will be an important tool in the human race's survival in the face of future pandemics. It's perhaps telling that when you try to find variants associated with getting critically ill with COVID-19 – an extremely well-powered examination of the effects of genetic variability on response to a pandemic disease! – the very largest effect sizes for individual variants are a doubling/halving of risk. This is a notable difference, but nowhere close to "total immunity" vs "certain death".

Having said that, I view "robust pathogen defence vs autoimmunity trade-off" as a very plausible just-so story, and likely to be true, but not concretely established at present. Sadly, it's one of those questions in science where running the right controlled experiment is practically impossible and we have to make do with detective work.

It is becoming increasingly clear that for many traits, the genetic effect sizes estimated by genetic association studies are substantially inflated for a few reasons. These include confounding due to uncontrolled population stratification, such as dynastic effects, and perhaps also genetic nurture^[1]. It is also clear that traits strongly mediated through society and behaviour, such as cognitive ability, are especially strongly affected by these mechanisms. 

You can avoid much of this confounding by performing GWAS on only the differences between siblings ("within-sibship GWAS") or between other pairs of family members ("family GWAS"). When you do this for cognitive ability, you find substantial deflation: heritability estimates decrease by around 45%, and the effect sizes estimated in population-level GWAS only correlate with these more direct effect estimates by about 0.55.

Does your analysis take this into account, for instance by using effect sizes estimated by within-sibship/family GWAS? If not, it would follow that genome editing would yield substantially lower increases in IQ than you estimate.

1. ^{^}

   Genetic nurture is a little complicated. The classic example in the GWAS context is parental genetic nurture. Here, you find an effect of a genetic variant on a trait in people, but the actual "direct" effect manifests only in those people's parents – one of whom must carry the variant, otherwise it would not be observed in the people in your study – and affected how they nurtured their kid, which then affected the trait you were measuring in their offspring. Genetically editing such a variant into an embryo would thus have no effect on that embryo when they are born and develop.

The "Black Death selection" finding you mention was subject to a very strong rebuttal preprinted in March 2023 and published yesterday in Nature. The original paper committed some pretty basic methodological errors^[1] and, in my opinion, it's disappointing that Nature did not decide to retract it. None of their claims of selection – neither the headline ERAP2 variant or the "half a dozen others" you refer to – survive the rebuttal's more rigorous reanalysis. I do some work in ancient DNA and am aware of analyses on other datasets (published and unpublished) that fail to replicate the original paper's findings.

1. ^{^}

   Some of the most glaring (but not necessarily most consequential): a failure to correctly estimate the allele frequencies underlying the selection analysis; use of a genotyping pipeline poorly suited to ancient DNA which meant that 80% of the genetic variants they "analysed" were likely completely artefactual and did not exist.