I think this is really cool. Is there any way I could get in touch with Gene Smith ?
Polygenic embryo screening is a commercially available service through which parents undergoing IVF can influence the traits of their future children. This can be used to decrease their risk of conditions like diabetes, heart disease various cancers, depression, schizophrenia and many others. All of the publicly available services focus on disease risk, and in particular on diseases of old age.
The service costs about $4000-$10,000 in addition to the cost of IVF.
There is one stealth mode company offering screening for non-disease traits such as intelligence and “externalizing” behavior, but their service costs substantially more. Anecdotally, I’ve heard it usually costs north of $15,000.
This is unfortunate because it means that only a select few unusually wealthy people will be able to afford the service.
In my view, there is no current reason why this should be the case; the data used to train the predictors for these traits is mostly openly available to the public, or at the very least to researchers through academic institutions.
I’ve put together a group to work on this project, with the goal of creating both an intelligence predictor and also generally a website where an arbitrary genome can be uploaded and scored according to any trait tracked by the PGS catalog.
We’ve made some early progress already; we have a working predictor which (according to some probably flawed validation test runs) was able to explain 12% of variance in intelligence. We still need to come up with a better method of validating the predictor, but we are making progress.
We’ve also created the beginnings of a website where users will be able to upload genomes and see polygenic scores for them.
With a tool like this, parents will be able to select embryos for other conditions besides just disease risk without paying an additional large extra fee.
The tool will also be useful for individuals who have used services like 23&Me to also assess their risk of different conditions.
I worked at Genomic Prediction for a year where I hired and ran the team that did research into the effects of diseases on life outcomes. I also collaborated with the data science team in charge of creating the predictors, and though I was never asked to create any predictors myself, I am familiar with the process and various techniques.
AG, the main person working on creating the predictor, has an extensive background in statistics and ML and has shown himself to be quite capable of creating predictors thus far.
Website work is being handled by users franzr and Jim Dandy. Franzr has limited experience building websites, but has learned quite quickly. Jim Dandy has web development experience, but like Franzr is using this project partially as a means of stretching his skills.
https://www.lesswrong.com/users/genesmith,https://twitter.com/GeneSmi96946389
At bare minimum we would need $1000 to pay for server costs, domain name registration, and other infrastructure. Additional money will allow us to pay members of the team, which I expect to considerably speed up rollout of the predictor. In fact, we may not get a predictor developed without this. Like most volunteer project, it has been quite difficult to keep people engaged and working on what are, at times, quite difficult problems. $10,000 would be enough to pay all of the contributors and to run servers for at least two years (assuming 10k users per year). So that is what I am requesting. If it is not available, we will do our best to continue working on the project anyways.
No response.
70%