The case for gengineering and my top SNP wishlist

With a quick primer on why we can't do this for adults yet

It’s a source of enduring frustration to me that we’ve had both the knowledge and the ability to do germline editing of humans for ~10+ years now, and it’s not legal or actually being done anywhere.

Yes, you can do embryo selection - I think it’s a great idea, and I hope you enjoy your 0.1 - 0.3 of a stdev bump, but that’s basically noise.

No, I’m talking about actual gengineering - in this particular case, getting in there and replacing one letter with another. Because that’s all you need for a lot of cool things! There are a TON of great SNP’s that could materially improve yours or your kids’ lives if this was available, and it’s entirely technologically feasible today (at least for kids).

Gengineering objections

And I know somewhere out there, there’s somebody gearing up to object to this, citing unknown unknowns, externalities, loss of genetic diversity, Red Queen’s Races, and eugenics.

Here’s why we should be able to do this to our kids:

Unknown unknowns and externalities:

There is today ZERO limitation on sickly, violent, unemployed, criminal, short time preference, or any other maladaptive combination of traits deciding to have a kid together, which has much larger and easier to quantify risks in the child and to society, but is seen as totally fine.

I argue that we should be able do this in the positive direction, because it's the norm and the standard today. Any two ill-advised people can decide to have kids and do it. Two people with Down's Syndrome, two criminals, whoever - pick any two people, they can make a kid today, and they're assuming those same risks, AND imposing costs on the rest of us.

Unknown unknowns? They have them. Externalities? They have those too, and they’re definitely negative.

It's a good bet that any externalities my gengineered kids will impose on society will be POSITIVE. But even if that's not the case, it should still be my decision to make, just like it is for every other parent in the world.

Loss of genetic diversity:

Not a real worry - any gengineering is going to A) be a tiny part of the world's wealthy doing it for at least the first 10 years, and B) we're going to have massive amounts of full-diversity "legacy" DNA around even if by some miracle 80% of the population were gengineering. We can sequence Neanderthal and Denisovan DNA, I don't think we have a lot to realistically worry about here.

Red Queen’s races around positional traits:

I mean take this to the extreme - the minimum standard to get into Harvard is now "6' 6" Olympic-medaling von Neumann adonis." Why would this be a BAD thing???

The fact that future generations are going to be attractive globe-straddling colossi in all fields of endeavor is an *unmitigated good,* that we should happily pull out our eyeteeth to achieve.

You're telling me you would be *disappointed* if your kid was an attractive, Olympic-medaling, von Neumann-level genius?? Or anywhere on the road between “average people now” and that?

This is eugenics / it will create rich people castes:

Pretty much anyone reading this is in finance, AI, startups, software, or are Professional Managerial Class. Rationalists and rat-adjacents are elites by pay, by IQ, and by occupation. And everyone in our circles optimizes hard on the quality of their mates - what is this, but eugenics?

I don't understand why when you suddenly use some science to go a bit farther, it's suddenly verboten to most people here. You already spent a *decade,* and a lot of effort, trying to optimize this to the n-th degree via dating!

Additionally, this is positive eugenics, where people are making choices about their mates and the kids they have, rather than negative eugenics, where people are being sterilized or killed.

Second, the rich ALREADY socially stratify, and are basically a caste! That's what The Son Also Rises is about. It's been true for thousands of years!

And finally, there's literally no path for "regular" people to get to a place where they can select away genetic defects and select into any benefits UNLESS you go through early adopters, ie the rich.

Big picture, economic growth is the strongest lever and driver for eliminating poverty worldwide (that’s what’s lifted ~1B people out of poverty in the last 40 years), and we will eventually be able to select on those things that will enable the elimination of poverty worldwide, ie. IQ and conscientiousness (among other traits).

The fact that somewhere along the way people will also be able to make their kids blonde and tall and healthy and strong and attractive as well (horrors! Aryan master race stuff!) is totally a personal choice for those parents, and is a GOOD thing.

From whatever perspective you come from: individual parents' rights and choices, societal impacts, or from the perspective of eliminating poverty and building a better future for the human race overall, gengineering is going to be one of the biggest levers we have to allow more choice, better societies, and less poverty. Why would we discard this positive eugenics tool, just because Nazi's did it in an explicitly biased, negative eugenics way in the past?

Psychopathic parents could breed kids for maximum conformity and use them as slaves:

Crazy I know, but somebody actually thought of this and brought it up to me once in a discussion about gengineering downsides. On the balance of civilizational and human suffering, which arm is bigger? The heart disease / diabetes / cancer / mental illness arm, currently affecting approximately everyone, or the "exceptionally psychopathic parents could create literal slaves for children" arm, which has basically nobody? Psychopathic parents can have as many kids as they want TODAY, and nobody does anything! There’s no parenting license! I’m sure laws would be written very quickly if this ever actually happened - arguably, gengineering would increase legibility and actually give us a way to stop psychopathic parents more often.

Whew. So let’s get down to the actual SNP’s that we know about!

Simple SNP’s that we can do basically tomorrow

All these are SNP’s - I have more traits in my wish list that require GWAS’s and suites of gene edits to achieve (in particular some of the big ones, like IQ, conscientiousness, robust mental health, and high happiness set point), but am leaving them out for this article because they’ll require massively parallel CRISPR capability.¹

SNP’s that require more calories, but have minimal downside

These are the low hanging fruit, and they are some of the sweetest fruit. It’s obvious why these didn’t reach fixation in the EEA. Needing 10%, 20%, even 5% more calories is a strong dealbreaker, especially given the many and varied climactic shifts, natural disasters, famines, and genetic bottlenecks we’ve gone through as a species.

But as far as I can tell in my research, and please correct me if you know or even suspect otherwise, essentially the ONLY cost to these is needing a little more calories. Which is a strict *advantage* in today’s world, where 75% of the US population is overweight or obese!

Myostatin knockouts - bully whippet, Belgian Blue, 7 month old German kid whose parents want anonymity, Nguyen Hoang Nam of Vietnam

• MSTN - myostatin knockout - The mutation behind bully whippets, belgian blues, and at least 2 humans, myostatin inhibits muscle growth, and knocking it out allows basically easy, “free” hypertrophy and lean mass.

• DEC2 - sleep 1-3 hours less SNP - The mutation changes a C to G in the DNA sequence of DEC2, which is predicted to cause a proline-to-arginine alteration at amino acid position 385 of DEC2. Noncarriers average 8.06 hours, and carriers average 6.25 hours. Carriers of these and similar DEC2 mutations also perform significantly better under sleep deprivation.²

  • Can you imagine how huge this is? Over a hundred year lifetime, this gives you back something like 7.5 YEARS of additional life! And that’s before we even get to GWAS studies and trying to optimize sleep further than an SNP!

• PEPCK-C overexpression - In muscle, this allows greatly improved metabolism and endurance (see 'mighty mice'). Synergistic effect with myostatin inhibition is untested, but intriguing.

Safety and risk reduction genes

• APOE replacement of APOE4 - APOE4 has notably higher Alzheimers incidence, notably higher risk of bleeding or damage after concussion, etc.

• BRCA breast cancer genes - you want to explicitly have the wild type and avoid the ~4 harmful mutations that increase risk.

• PCSK5 and PCSK9 - Those with the PCSK5 gene have 88 percent lower coronary disease. PCSK9 (according to George Church) is also capable of lowering susceptibility to coronary disease.

• FUT2 - Those with double FUT2 are resistant to norovirus / stomach flu.

• Protective mutations for heart attacks:

  PCSK9 - 80% lower risk

  NPCILI - 53% lower risk

  LPA - 24% lower risk

  APOC3 - 40% lower risk

  ANGPTL3 - 34% lower risk

  ANGPTL4 - 53% lower risk

  ASGR1 - 34% lower risk

Athletics genes

• 21 HERITAGE study genes - the HERITAGE study looked at genetic variants in terms of exercise response, and highlighted variants correlated with higher and lower response,³ which leads to significant VO2max, strength, and cholesterol differences.

• LRP5 G171V/+, extra strong bones - great for any athlete, but particularly power athletes.

• ACE and ACTN3 - Fast twitch vs endurance proportion genes, you can tune either way.

• EPO receptor SNP upgrade - As seen in 7x Olympic medalist Eero Mäntyranta and his extended family, an adenine→guanine swap in one (but not both) of the EPO receptor genes, which stops the receptor construction about 85% of the way through and makes the receptor respond maximally to even minimal bloodstream EPO. The family members carrying the mutation actually have longer lives (as well as Eero having much better athletic performance than baseline) than the ones without.

Cosmetic edits and general “winning” corrrelates

• rs7495174 - green eye color.

• rs12913832 - blue eye color.

• MC1R rs1805005, known as Val60Leu or V60L, associated with light blond hair in one study (PMID 9302268).

• IRX3 and IRX5 - Anti-obesity SNP - in FTO region, Switching the C to a T in obesity risk individuals turned off IRX3 and IRX5, restored thermogenesis to non-risk levels, and switched off lipid storage genes.⁴

• Height SNP's - SNP near GDF5-UQCC locus (e.g., rs143383), SNP near FTO gene (e.g., rs8050136), SNP near HMGA2 gene (e.g., rs1042725) - these three are the three biggest that we know about, and can drive 0.3-0.6cm each.

• ABCC11 -/-, low odor production - prevalent in Asian populations, people with this variant have low body odor.

• rs10776614-T - Self employment. There’s another variant that shows up in a GWAS, but it’s also strongly associated with obesity, drug addiction, etc., presumably from maximizing “IDGAF” / lowering agreeability.

• Extroversion - rs2164273-G - gives you both a slight bump towards extroversion and a slight bump away from higher neuroticism, rs57590327-T - another slight bump to extroversion.

• High number of children - rs13161115-C, rs10908474-A, rs2415984-A, rs10009124 - because what’s the POINT of kids? Grandkids!

Alright, so this was a fun little survey of some interesting things we could do basically tomorrow if every government wasn’t a bunch of regressive, stuck-up prudes.

Obviously we know nothing about interaction effects for these without some major GWAS’s and number crunching, in particular because many of them are relatively rare variants. So the final menu you would choose is really down to your risk appetite.

I would personally probably err on the higher side, in the sense that if the world was civilized and I could pay to put these things into my kids, I’d probably end up putting 10-17 changes into them. This is because often, all good things are associated with each other (like IQ being positively associated with things as diverse as looks, good health, and reaction speed), so I think the odds of catastrophic interactions are relatively low. But I would research what we know about protein interactions and metabolic chains pretty well before each decision, including having geneticists and proteomics pros chiming in.

What are our best bets for actually getting effective gengineering?

Honestly, Prospera and the like. There are some signs that Japan is more open to gengineering, because uniquely they allow human germline edits for research (which no other developed country does to their extent).

I’d personally bet on Thailand being a pretty good candidate eventually, but think that’s probably 10+ years down the road, because it’s not even on the radar there right now. Ukraine and Russia both currently allow limited human germline edits, too, and it’s at least plausible that the war aftermath could push either one in a positive direction.

For now, some break through in iterated embryo selection or meiosis is probably our best bet.

Now, the REAL question - why can’t we put any of these into ourselves? Particularly for the sleep SNP! Man, I would happily pay more than a million bucks just for that one alteration alone!

Why we can’t do this to adults

Broadly, you have ~37 trillion cells, and spreading a genetic edit to all the relevant cells in a timely way is basically impossible. So what’s the main obstacle here? Transmissibility. Penetration. Incidence. Getting it into the trillions of cells it needs to get into in some reasonable amount of time (and without inspiring immune reactions, and while targeting with sufficient fidelity that the changes are reliably made).

The current state of the art for in-vivo gengineering operations is adeno-associated viruses, lentiviruses, or lipid nanoparticles.

Adeno associated viruses - these are good because they have relatively limited pathogenicity, and have the ability to stably integrate into the host cell genome at a specific site (designated AAVS1) in the human chromosome 19. You can only put small genes in via AAV’s (2-4kb), and many people have immune responses due to being infected with the wild type at some point, which can attenuate or block any actual changes.

Lentiviruses - a retrovirus derived from HIV, which can carry up to 8-10kb of genetic info, although 3-4kb is optimal. Uniquely, they can infect non-dividing cells, not just dividing cells, which is a pretty big deal in terms of penetration. They generally induce fewer immune reactions than AAV’s. However, they’re complex and difficult to manufacture, and can’t be scaled currently.

Lentivirus GFP-insert mice bracketing a normal mouse

Lipid nanoparticles - we can manufacture these at scale, and with a CRISPR-cas9 payload, they can create stable genetic changes. They can only hold a 1-5kb payload after CRISPR-cas9 is loaded, and they’re a little harder to target. They generally don’t directly activate immune responses, but can cause inflammation or allergic reactions.

One bit of good news with all these kb limitations - SNP changes are only 1-2kb in general (the actual SNP change is 0.1kb, but the infrastructure around it like HDR to drive the change is 1-2kb).

Real-world results:

Jennifer Doudna is the Nobel-winner and originating PI when it comes to CRISPR-cas9 editing, and her company Editas Medicine did one of the most ambitious real-world tests, using adenoviruses to try to correct a mutation in the CEP290 gene that causes blindness. Although it was safe and didn’t have negative side effects, there were only minor visual gains. Something went wrong, and they’re digging into what.

Lipid nanoparticles successfully treated Transthyretin Amyloidosis, a disease involving a misfolded protein that builds up amyloids causing cardiac myopathy and heart failure with high mortality (generally fatal within 2-6 years). A one-time infusion into the bloodstream with RNA targets for the relevant single gene did significantly lower levels of the relevant protein, and the levels remained at that lower level for the entire 4-6 months of followup. The latest news is that they recruited 765 people to a Phase III trial, which given the incidence of the disease (~1 in 100-160k, or ~2 - 3.5k in the entire country, many of whom will not be symptomatic, so they likely got a majority of symptomatic people in the entire country into this Phase III trial), is a fairly strong and credible signal that they believe this is a cure.

You’ll note the two attempts had relatively conservative targets which greatly limited the amount of cells needing to be altered (retinas and livers), so broadly systemic changes like myostatin or sleep SNP’s are still VERY far away for adults, and we may honestly never get there.

So for now, we’re stuck with only being able to alter things for the next generation.

I hope this has inspired some interest and some thoughts in you, and maybe pushed you a little bit towards seeing gengineering as a good thing, and entirely in line with the leeway and affordances we give to ALL parents today.

1. Which in my opinion, we should be funding crash, “Space Race” style programs around, to try to make effective gengineering a reality, because then we’ll be able to alter the necessary tens, hundreds, or thousands of genes to drive significant deltas in IQ, conscientiousness, high happiness set point, lower neuroticism, etc.

2. For the “sleep less” variant mentioned, see https://pmc.ncbi.nlm.nih.gov/articles/PMC2884988/, for the other variants, which optimizes NREM sleep at the cost of REM sleep and better performance under sleep deprivation, see https://pmc.ncbi.nlm.nih.gov/articles/PMC4096202/

There are some other candidates too - NPSR1 - Mutant neuropeptide S receptor reduces sleep duration with preserved memory consolidation. And BHLHE41 Y362H - rs121912617 (another DEC2 mutation) is also associated with short sleep and resistance to sleep deprivation.

3. HERITAGE study breakdown:

WANT:

-634 G>C VEGFA variant at 6p21.1 - VO₂max gain ~10–15%.

ACE insertion at 17q23.3 - 0.16 mL/kg/min per allele (VO₂max gain).

Arg16Gly, Gln27Glu ADRB2 variants at 5q31-32 - Gly16 allele linked to greater improvement in VO₂max; Glu27 linked to better fat metabolism. Gly16: 5–10% difference in VO₂max gain

Trp64Arg ADRB3 variant - Arg64 allele linked to reduced fat oxidation and insulin sensitivity improvements.

E40K (Glu40Lys) ANGPTL4 variant at 19p13.2 - Lys40 allele linked to better lipid profile responses to exercise, 10% change in triglyceride levels

-75 G>A APOA1 variant at 11q23.3 - A allele associated with ~5-10% greater HDL increase following exercise.

C825T GNB3 variant at 12p13.31 - T allele associated with 3-5mmHG improved blood pressure response to exercise.

Asp358Ala (rs2228145) IL6R variant at 1q21.3 - Ala358 associated with ~10-20% CRP reduction effects post-exercise.

FokI (rs2228570) VDR variant at 12q13.11 - ~10% strength improvement.

Pro12Ala PPARG variant at 3p25.2 - ~15% improvement in insulin sensitivity.

Ser447Ter LPL variant at 8p21.3 - Triglyceride reduction ~15%.

-514 C>T LIPC variant at 15q21.3 - HDL-C increase ~5%.

R577X ACTN3 variant at 11q13.2 - XX genotype with lower power output but better endurance, with up to 10% endurance variation

AVOID:

G34T (C34T) AMPD1 variant at 1p13.2 - T allele associated with ~5% reduced VO₂max response to exercise.

T2488G APOB variant at 2p24.1 - G allele linked to lower LDL improvements post-exercise.

-174 G>C IL6 variant at 7p15.3 - VO₂max gain ~5% lower with G allele.

Gly482Ser PPARGC1A at 4p15.2 - VO₂max gain ~5% lower.

-308 G>A TNF variant at 6p21.3 - VO₂max gain ~5% lower.

4. And honestly with ~75% of the US overweight or obese, aren’t we ALL in the “obesity risk” tranche at this point?

The researchers predicted that a genetic difference of only one nucleotide is responsible for this obesity association. In risk individuals, a thymine (T) is replaced by a cytosine (C) nucleobase, which disrupts repression of the control region and turns on IRX3 and IRX5. This then turns off thermogenesis, leading to lipid accumulation and ultimately obesity.

By editing a single nucleotide position using the CRISPR/Cas9 system — a technology that allows researchers to make precise changes to a DNA sequence — the researchers could switch between lean and obese signatures in human pre-adipocytes. Switching the C to a T in risk individuals turned off IRX3 and IRX5, restored thermogenesis to non-risk levels, and switched off lipid storage genes.