> But the idea here is that, by checking thousands of genes, you can make predictions for things that start to be very relevant to parents, like attractiveness or height or intelligence.
Personally, I think this interpretation of polygenic risk scoring is a crock of irreproducible shite. I don’t think it’s your fault, I think a lot of people in the field are shilling something they don’t have for grant money, and they’ve created an exciting sci-fi yarn that’s easy for people outside the field to digest.
Height is a univariate trait that does not change after the age of ~20. If you show me someone who can take a genome and predict adult height within a centimeter, then I’ll believe they have a snowball’s chance of predicting something as nuanced and varied as “attractiveness” or “intelligence”.
I think there’s promise in applying PRS to assessing risk for non-Mendelian disease, but there’s far too many social and environmental variables at play to reliably predict these softer features. Like other people have joked, if you want a single number that best predicts educational attainment, use a zip code.
> show me someone who can take a genome and predict adult height within a centimeter, then I’ll believe they have a snowball’s chance of predicting something as nuanced and varied as “attractiveness” or “intelligence”.
This has been done. Not to 1cm but ~3cm, and it has been reproduced by many labs all over the world
That said, most genetic prediction aren't targeting exact vales, but avoiding the worst outcomes, which is a lot easier.
Instead of saying this embryo will go grow to X cm, the claim is this embryo is 90% likely to grow to be taller than average.
Same for IQ. They don't predict the IQ, but decrease poor outcomes and increase high ones. Hell, we have had rough monogenic screening for IQ since the 1960s
I haven't previously read the Lello paper [1] that review is citing, but a quick skim makes me skeptical.
Particulary the legend for figure 5:
> Activated SNPs are distributed roughly uniformly throughout the genome.
If the authors were actually identifying a genetic component to a heritable trait, I'd expect them to observe some linkage disequilibrium. And without any analysis of the SNPs (are they coding/noncoding? which genes are they associated with?) it's hard to believe that they're uncovering actual biology and not just chance correlates with external/socioeconomic factors.
I also find it difficult to trust studies when the lead author fails to disclose a conflict of interest. [2]
> it's hard to believe that they're uncovering actual biology and not just chance correlates with external/socioeconomic factors.
This is routinely validated on siblings. If the polygenic scores are as accurate in predicting differences between siblings (who, presumably, share substantially same environment) as they are between unrelated people, it means that they detect real, biological things, instead of just some kind of population stratification. The linked abstract, of course, mentions this, and you'd have known this if you had read the article.
Also, your assumption that socioeconomic factors are independent of genomes (that they are just "chance correlates) is also substantially wrong. Genes correlate with socioeconomic factors, because they often cause socioeconomic factors. People are not born into socioeconomic conditions randomly, they are born into socioeconomic conditions of people who share half of their genome with them.
Understanding whether the model is capturing biology is critical when thinking about applying it to IVF. If the model is primarily capturing socioeconomic correlates, those factors will (in most cases) be fixed for all embryos from a given pair of parents. The PRS needs to be weighting _biological_ risk conditioned on a fixed environment if its to be used ethically in this context.
Nature can still be important for even highly heritable traits. For example, the size of a person's vocabulary is highly heritable, but a feral human will have very limited vocabulary. Such traits still have a very strong biological basis.
The control on those are in places that have relatively standard nutritional distribution. The environment obviously plays a strong role, just look at the height difference in populations between north and South Korea.
> Height is a univariate trait that does not change after the age of ~20. If you show me someone who can take a genome and predict adult height within a centimeter, then I’ll believe they have a snowball’s chance of predicting something as nuanced and varied as “attractiveness” or “intelligence”.
I don't know why that would be the standard, since the amount of variation in height that's attributable to environmental factors might make that fundamentally impossible. But we can get quite close to that standard nonetheless.
And even if it's not possible yet, do you think that in 20 years we'll have no ability to predict height or intelligence
from a genome? It seems very plausible to me, especially with how cheap genome sequencing is now.
> since the amount of variation in height that's attributable to environmental factors might make that fundamentally impossible
Yes, that is my point.
Both intelligence and attractiveness have a significant amount of variation attributable to environmental or other external factors, and have the additional complication that they cannot be measured by a single objective unit (like height can).
> do you think that in 20 years we'll have no ability to predict height or intelligence from a genome
The quantification of intelligence is notoriously confounded by socioeconomic factors. I do not think talking about predicting a feature makes sense while we are currently unable to describe it well.
IMO dedicating funding to improving child care, healthcare+diet, and k-12 education will have a much greater impact on increasing a society's measures of intelligence and educational attainment. There's much stronger evidence that these factors are associated with improved outcomes. But, the work isn't "sexy" and doesn't come with a sci-fi flair.
Kind of like ignoring climate work in favor of Mars colonization. There seems to be a cultural bias in tech towards moonshot panaceas vs doing the unglamorous grind. It makes me think of the Bill Gates quote that "a lazy person will find an easy way to do a hard job," and while that's valuable in some contexts I don't think it's universally applicable.
> The quantification of intelligence is notoriously confounded by socioeconomic factors. I do not think talking about predicting a feature makes sense while we are currently unable to describe it well.
If we are doing this selection of IVF embryos it is feasible to control for these confounding factors.
> IMO dedicating funding to improving child care, healthcare+diet, and k-12 education will have a much greater impact on increasing a society's measures of intelligence and educational attainment. There's much stronger evidence that these factors are associated with improved outcomes. But, the work isn't "sexy" and doesn't come with a sci-fi flair.
Source? Isn't intelligence in adulthood more correlated with your parents than any of these environmental factors? That was at least what I had recalled from twin adoption studies.
> IMO dedicating funding to improving child care, healthcare+diet, and k-12 education will have a much greater impact on increasing a society's measures of intelligence and educational attainment.
Maybe if you do it in North Korea, but US/Europe are already pretty much "maxed out" here. You can always improve things on the margin, but you're not going to see substantial gains of even d = 0.5 magnitude.
Have you applied the same statistical rigor to the evidence that improved child care, healthcare+diet, and k-12 education improve intelligence and educational attainment leads to intelligence increases as you do for genetic arguments?
Personally, I think this interpretation of polygenic risk scoring is a crock of irreproducible shite. I don’t think it’s your fault, I think a lot of people in the field are shilling something they don’t have for grant money, and they’ve created an exciting sci-fi yarn that’s easy for people outside the field to digest.
Height is a univariate trait that does not change after the age of ~20. If you show me someone who can take a genome and predict adult height within a centimeter, then I’ll believe they have a snowball’s chance of predicting something as nuanced and varied as “attractiveness” or “intelligence”.
I think there’s promise in applying PRS to assessing risk for non-Mendelian disease, but there’s far too many social and environmental variables at play to reliably predict these softer features. Like other people have joked, if you want a single number that best predicts educational attainment, use a zip code.