Yeah, it's funny like that. I did see it a few more places, now, e.g. here
bsky.app/profile/dr-a...

- but it seems it didn't really "catch"

Godt de ikke også kan fejre det samme for 10-måneders året!

And finally, may I remind everyone that the goal is, and always was, to lower the bar for access to genetic information that can improve the health of everyone.

Origin | Nucleus Labs A family of the most predictive genetic models ever built.

Here's a link to where the scores can be downloaded
mynucleus.com/labs/origin

Here's a link to Stephan's more detailed point by point response on the specific scientific allegations
x.com/stephan_cdgn...

We welcome discussion, but right now it seems many on X are just repeating incorrect and planted claims that can be easily verified independently. Besides, I note that the we apparently are the only ones in this debate that chose to release our full scores.

But it feels overwhelming to see these massively popular Twitter accounts seek to distort our words and intentions and I'm sure we have not seen the last of it. I would therefore encourage those of you in my network that have the time and ability, to join in on independently scrutinizing this work

So far the factual errors found are:
* A typo of an ICD-9 code in a supplementary table
* A sentence with a typo that wrongly muddles the distinction between training and validation sets
Those are trivial and will of course be corrected.

It is true that we have used the same biobanks, UK Biobank, All-Of-Us, but who has not? We have also used the same SBayesRC method developed by researchers at the University of Queensland, for calculating the weights. It seems to be the current state of the art, why would we not use it?

I want to assure all my friends and peers that of course it is not fraud and plagiarism. We have been running polygenic risk scoring for more than a decade now, and this is simply the latest step in a long trail of ever-improving predictive algorithms.

However, it seems to have riled up the people in and around the IQ-prediction company Herasight, so much that we have now gotten their entire fanbase in the IQ-genetics crowd screaming about fraud and plagiarism and worse.

I don't think I've ever had a pre-print more heavily scrutinized than the one we recently put on medRxiv with the new Nucleus Genomics PRS for 9 diseases:
www.medrxiv.org/content/10.1...
This is great! We welcome discussion!

Yeah, agreed. I'd definitely definitely also recommend to focus on things like cancer PRS, and then only after classic/rare mendelian effects have been excluded. Maybe also Alzheimer's, diabetes and cardiovascular disease - the predictive power of PRS for those is quite good nowadays.

When there's several viable, euploid, embryos without severe mendelian disorders left to choose from, I honestly don't see the problem of looking at polygenic risk scores among the remaining. Some of them can really reduce the disease burden a lot.

what makes you think people wouldn't select for viability, e.g. euploidy, first?

Nationwide, Couple-Based Genetic Carrier Screening - PubMed Couple-based reproductive genetic carrier screening was largely acceptable to participants and was used to inform reproductive decision making. The delivery of screening to a diverse and geographically dispersed population was feasible. (Funded by the Medical Research Future Fund of the Australian g …

h63d is just exactly in, at the very least severe end of the spectrum - you are correct about that, but it's not what solely drive the 90%, far from it. Most people do have other and more severe recessive variants -
pubmed.ncbi.nlm.nih.gov/39565987/

Very proud of this work: State of the art polygenic risk scores, and open-access even. Really hoping that this can push the field forward towards better disease prediction and prevention. The first author, Stephan, is only on X - but he did a great explainer thread there too. Check it out.

Life will find a way!

Such a nice iniative.

Sex, age, cholesterol, blood pressure pretty good too. But they are all independent and can be used together, so there is that.

I don't think I am.

Probably because he or she followed me, and I just pressed follow back, idk. Have no idea who it is.

Look, I will be very happy to debate ethics with you, but I think you come off as slightly threatening here. What gives?

Following someone on bsky (or elsewhere) does not imply a recommendation of their opinion, in my book.

Yes, I understand that you think the spread of predictions should be more tight. That's why I ask what threshold you'd consider as ok? The reason I ask, is of course that some of the PRS are actually stronger predictors than PGT-M, which I assume you find to be ok.

That's fine! Thanks for discussing.

But it's not nothing either, and because of that I think it's a real problem that otherwise well-intended reporting guidelines inadvertently can lift it up to a status of strong but forbidden.

These are typical values btw, breast cancer PRS are very strong and it seems odd to me to not use that, if monogenic PGT-M is accepted as ok. IQ prediction strength is weak at best, few points at the max.

My view then, is to take openness of information to it's completion and instead emphasise accurate communication. Using absolute scales helps, e.g. comparing between +1 IQ and -30%points breast cancer risk frames the choices in a way that reflects the relative differences in prediction strength.

Yes, very difficult to draw a firm line, agreed, exactly. "PRS vs monogenic" won't work, because some PRS are stronger. Similarly "traits vs diseases" also can run into problems because extremes of traits often become diseases (e.g. dwarfism/height, mental-retardation/IQ).

How do they work then?