I hate to be the bearer of bad news, but the methodology in the study you linked is just terrible.
99% of candidate gene studies do not replicate, and the first study you link uses an extremely error-prone P < 0.05 threshold to determine that orexin has an effect.
It may still have an effect, but there’s no way to really know this without some kind of large sample GWAS.
The proper way to do this would be to get research access to a large genetic database like 23&Me or UK BioBank and pair that data with info from sleep trackers like Apple Watch, FitBit, or other devices to see which genes are involved in sleep duration and how large the effect of each is.
Depending on the sparsity of the trait, you’d probably need somewhere between 100k − 1 million genotype-phenotype pairs to make a good predictor.
Lastly, in the “Possible Actions” section you should include embryo selection. Genes obviously play some role in sleep needs. If at least some of them have no relevant downsides, selecting for those genes in embryos would probably be quite cost-effective. In fact, you could further reduce the odds of any unintended side-effects by selecting not just for shorter required sleep duration, but also for longer quality-adjusted lifespan.
When doing gene studies if you look at all genes and see which one’s have P < 0.05, that gets you a lot of false positives. That’s however not what they did in the linked paper.
The linked paper looks at one gene because previous papers identified that gene as being significant in humans. The paper says that the gene has significance for sleep in mice.
The argument doesn’t rest on a single experiment. I agree that it would be desirable to have 23&Me cooperate with Apple or Google to find genes that affect the metrics that the Apple watch / Fitbit measures. That’s helpful to have a good overview of all the mutations that affect sleep length.
Lastly, in the “Possible Actions” section you should include embryo selection.
Yes, that’s a valid action. I was thinking about actions that actually result push the field forward. You could have benefits for the child from such embryo selection but I wouldn’t expect it to lead to knowledge generation.
Can you link the earlier studies showing the significance of BHLHE41 in humans? When I Google it all I find are other candidate gene studies with small sample sizes.
https://www.science.org/doi/abs/10.1126/science.1174443 seems to be the first paper. It does have a small sample size and thus is only able to produce candidates, but it results in the later paper not randomly searching over all possible gene mutations.
My main point is that there are papers that made independent observations and thus arguments that a single paper doesn’t demonstrate the effect doesn’t hold. I didn’t copy the exact minute numbers that the EA cause report had because I was unsure about the exactness of the data.
In the case of orexin, my argument doesn’t just rest on the DEC2 gene.
The experiments that improved performance in sleep-deprived rhesus monkeys happened before the discovery of the link between the DEC2 mutation and orexin.
The observations in Astyanax mexicanus seem independent from my perspective. Attempts to make Astyanax mexicanus a model organism aren’t driven by sleep researchers but because it’s interesting for studying evolution.
Orexin deficiency causing Narcolepsy type 1 is independent of any findings about the DEC2 gene as well.
As far as the linked post of Scott goes, it says nothing about experiments on animals other than humans. Gene knockout studies in mice and Drosophila seem to me like a pretty good way to measure the influence of a gene.
I hate to be the bearer of bad news, but the methodology in the study you linked is just terrible.
99% of candidate gene studies do not replicate, and the first study you link uses an extremely error-prone P < 0.05 threshold to determine that orexin has an effect.
It may still have an effect, but there’s no way to really know this without some kind of large sample GWAS.
The proper way to do this would be to get research access to a large genetic database like 23&Me or UK BioBank and pair that data with info from sleep trackers like Apple Watch, FitBit, or other devices to see which genes are involved in sleep duration and how large the effect of each is.
Depending on the sparsity of the trait, you’d probably need somewhere between 100k − 1 million genotype-phenotype pairs to make a good predictor.
Lastly, in the “Possible Actions” section you should include embryo selection. Genes obviously play some role in sleep needs. If at least some of them have no relevant downsides, selecting for those genes in embryos would probably be quite cost-effective. In fact, you could further reduce the odds of any unintended side-effects by selecting not just for shorter required sleep duration, but also for longer quality-adjusted lifespan.
When doing gene studies if you look at all genes and see which one’s have P < 0.05, that gets you a lot of false positives. That’s however not what they did in the linked paper.
The linked paper looks at one gene because previous papers identified that gene as being significant in humans. The paper says that the gene has significance for sleep in mice.
The argument doesn’t rest on a single experiment. I agree that it would be desirable to have 23&Me cooperate with Apple or Google to find genes that affect the metrics that the Apple watch / Fitbit measures. That’s helpful to have a good overview of all the mutations that affect sleep length.
Yes, that’s a valid action. I was thinking about actions that actually result push the field forward. You could have benefits for the child from such embryo selection but I wouldn’t expect it to lead to knowledge generation.
Can you link the earlier studies showing the significance of BHLHE41 in humans? When I Google it all I find are other candidate gene studies with small sample sizes.
https://www.science.org/doi/abs/10.1126/science.1174443 seems to be the first paper. It does have a small sample size and thus is only able to produce candidates, but it results in the later paper not randomly searching over all possible gene mutations.
My main point is that there are papers that made independent observations and thus arguments that a single paper doesn’t demonstrate the effect doesn’t hold. I didn’t copy the exact minute numbers that the EA cause report had because I was unsure about the exactness of the data.
I’m not sure if you have read the story of 5HTTLPR and all the independent studies which found it to have an effect, but if you haven’t you should.
In the case of orexin, my argument doesn’t just rest on the DEC2 gene.
The experiments that improved performance in sleep-deprived rhesus monkeys happened before the discovery of the link between the DEC2 mutation and orexin.
The observations in Astyanax mexicanus seem independent from my perspective. Attempts to make Astyanax mexicanus a model organism aren’t driven by sleep researchers but because it’s interesting for studying evolution.
Orexin deficiency causing Narcolepsy type 1 is independent of any findings about the DEC2 gene as well.
As far as the linked post of Scott goes, it says nothing about experiments on animals other than humans. Gene knockout studies in mice and Drosophila seem to me like a pretty good way to measure the influence of a gene.