The lack of good population genetic information in animal models and deep phenotyping of complex behavioral traits is probably one of the biggest impediments to robust animal testing of this general approach.
(For people reading this thread who want an intro to finemapping this lecture is a great place to start for a high level overview https://www.youtube.com/watch?v=pglYf7wocSI)
Kind of, there are many ways that changed in isolation get you a bit more oxygen but many of them act through the same mechanism so you change 1000 things that get you +1 oxygen on their own but in combination only get you +500.
To use a software analogy imagine an object with two methods where if you call either of them a property of an object is set to true, it doesn't matter if you call both methods or if you have a bunch of functions that call those methods you still just get true. Calling either method or any function that calls them is going to be slightly correlated with an increased probability the the property of the object will be true but it does not add. There are many way to make it true but making it true more times does not make it 'more true'.
If we change this from a boolean to an integer then some methods might only increment it if it is not already greater than some value specific to the method.
I'm curious about the basis on which you are assigning a probability of causality without a method like mendelian randomisation, or something that tries to assign a probability of an effect based on interpreting the biology like a coding of the output of something like SnpEff to an approximate probability of effect.
The logic of 30% of its effect based on 30% chance it's causal only seems like it will be pretty high variance and only work out over a pretty large number of edits. It is also assuming no unexpected effects of the edits to SNPs that are non-causal for whatever trait you are targeting but might do something else when edited.
Could you expand on what sense you have 'taken this into account' in your models? What are you expecting to achieve by editing non-causal SNPs?
The first paper I linked is about epistasic effects on the additivity of a QTLs for quantitative trait, specifically heading date in rice, so this is evidence for this sort of effect on such a trait.
The general problem is without a robust causal understanding of what an edit does it is very hard to predict what shorts of problem might arise from novel combinations of variants in a haplotype. That's just the nature of complex systems, a single incorrect base in the wrong place may have no effect or cause a critical cascading failure. You don't know until you test it or have characterized the system so well you can graph out exactly what is going to happen. Just testing it in humans and seeing what happens is eventually going to hit something detrimental. When you are trying to do enhancement you tend to need a positive expectation that it will be safe not just no reason to think it won't be. Many healthy people would be averse to risking good health for their kid, even at low probability of a bad outcome.
There are a couple of major problems with naively intervening to edit sites associated with some phenotype in a GWAS or polygenic risk score.
The SNP itself is (usually) not causal
Genotyping arrays select SNPs the genotype of which is correlated with a region around the SNP, they are said to be in linkage with this region as this region tends to be inherited together when recombination happens in meiosis. This is a matter of degree and linkage scores allow thresholds to be set for how indicative a SNP is about the genotype a given region.
If it is not the SNP but rather something with which the SNP is in linkage that is causing the effect editing the SNP has no reason the effect the trait in question.
It is not trivial to figure out what in linkage with a SNP might be causing an effect.
Mendelian randomisation (explainer: https://www.bmj.com/content/362/bmj.k601) is a method that permits the identification of causal relationships from observational genetic studies which can help to overcome this issue.
In practice epistatic interactions between QTLs matter for effects sizes and you cannot naively add up the effect sizes of all the QTLs for a trait and expect the result to reflect the real effect size, even if >50% effect are additive.
Terminology:
epistasis - when the effect of a genetic variant is dependent on the genotype of another gene or genes to have an effect.
QTL - quantitative trait locus, a location in the genome where the genotype is correlated with a quantitative phenotype e.g. height
A hypothetical example of how epistasis can lead to non-additivity in QTLs:
SNPs linked with genes A, B and C are associated with some trait.
Variant A is more a more active kinase than regular A that phosphorylates and activates C, So is variant B
Phosphorylation of C is effectively binary, if either A or B does it does not matter so editing either has the same effect.
Variant C is active even when not phosphorylated so editing A and/or B has no effect beyond that of editing C - except maybe side effects from now phosphorylating something else.
In agronomy where this has been best studied with the goal of engineering crops with specific complex traits once you start trying this epistatic effects show up.
For example:
https://doi.org/10.1038/s41598-018-20690-w
https://doi.org/10.1007/s00122-010-1517-0
The (much) bigger problem is not editing a bunch of bases in the embryo it's knowing which ones to edit (safely).
I have a reading list and recommendations for sources of ongoing news in the longevity/immortality/life extension space in the show notes for the recent special episode of my podcast where my co-host Michael and I discuss ageing and immortality. We are both biology PhDs, my background is in the epigenetics of ageing and Michael's bone stem cells.
https://www.xenothesis.com/53_xenogenesis_ageing_and_immortality_special/
I should add "Immune: a Journey into the Mysterious System that Keeps You Alive" to that list actually.
In particular from that list I recommend these for 'coming at biology from a physics perspective':
To clarify it's the ability to lock you're bootloader that I'm saying is better protection from 3rd parties not the propriety nature of many of the current locks. The HEADs tools for example which allows you to verify the integrity of your boot image in coreboot would be a FOSS alternative that provides analogous protection. Indeed it's not real security if it's not out there in the open for everyone to hammer with full knowledge of how it works and some nice big bug bounties (intentional or unintentional) on the other side to incentivise some scrutiny.
Thanks for the link. The problem of how to have a cryptographic root of trust for an uploaded person and how to maintain an on going state of trusted operation is a tricky one that I'm aware people have discussed. Though it's mostly well over my cryptography pay grade. The main point I was trying to get at was not primarily about uploaded brains. I'm using them as an anchor at the extreme end of a distribution that I'm arguing we are already on. The problems of being able to trust its own cognition that an uploaded brain has we are already beginning to experience in the aspects of our cognition that we are outsourcing.
Human brains are not just general purpose CPUs much of our cognition is performed on the wetware equivalent of application-specific integrated circuits (ASICs). ASICs that were tuned for applications that are of waning relevance in the current environment. They were tuned for our environment of evolutionary adaptiveness but the modern world presents very different challenges. By analogy it's as if they were tuned for sha256 hashing but Ethereum changed the hash function so the returns have dropped. Not to mention that biology uses terrible, dirty hacky heuristics that would would make a grown engineer cry and statisticians yell WHY! at the sky in existential dread. These leave us wide open to all sorts of subtle exploits that can be utilised by those who have studied the systematic errors we make and if they don't share our interests this is a problem.
Note that I am regarding the specifics of an uploaded brain as personal data which should be subject to privacy protections (both at the technical and policy level) and not as code. This distinction may be less clear for more sophisticated mind upload methods which generate an abstract representation of your brain and run that. If, however, we take a conceptually simpler approach the data/code distinction is cleaner. let's say we have an 'image' of the brain which captures the 'coordinates' (quantum numbers) of all of the subatomic particles that make up your brain. We then run that 'image' in a physics simulation which can also emulate sensory inputs to place the uploadee in a virtual environment. The brain image is data, the physics and sensory emulation engine is code. I suspect a similar reasonable distinction will continue to continue to hold quite well for quite a while even once your 'brain' data starts being represented in a more complex data structure than and N dimensional matrix.
I actually think mind uploading is a much harder problem than many people seem to regard it as, indeed I think it is quite possibly harder than getting to AGI de novo in code. This is for reasons related to neurobiology, imaging technology and computational tractability of physics simulations and I can get into it at greater length if anyone is interested.
GeneSmith gave some more details about his background in this episode of the Bayesian Conspiracy podcast: https://www.thebayesianconspiracy.com/2025/02/231-superbabies-with-gene-smith/