{"items": [{"author": "Jess", "source_link": "https://www.facebook.com/jefftk/posts/972501387302?comment_id=972518982042", "anchor": "fb-972518982042", "service": "fb", "text": "> This caveat about stacking is pretty serious, though.

As you probably know, there are arguments, both theoretical and empirical, that SNP risk factors for polygenic traits generally \"stack\" (and are, by definition, much smaller than 2x), e.g., Steve Hsu. But even though most traits are polygenic (under some reasonable partitioning of an organism's complete phenotype into discrete traits) they are also the hardest to learn the genetic basis of because they require genotype-trait samples from many measured individuals to tease out of the data. Therefore, the traits for which we actually understand the genetics are most likely to be simple (e.g., Mendelian), and the above arguments don't apply.

In your report, how many individual SNPs are generally contributing? Like just 2 or 3? Or are there really examples of a dozen SNPs all contributing >2x to a single trait?", "timestamp": "1544894525"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/972501387302?comment_id=972518982042&reply_comment_id=972528937092", "anchor": "fb-972518982042_972528937092", "service": "fb", "text": "→ Let's take baldness, an arbitrary and non-distressing trait for which Promethease lists a good number of relevant SNPs. It gives 10 SNPs marked as \"bad\" and 10 marked as \"not set\". For the ones marked as \"bad\" magnitude runs from 1.1 to 3.1, while for the \"not set\" ones magnitude runs from 0.5 to 1.0.

The \"bad\" ones are:

1. 7x: https://www.snpedia.com/index.php/gs122
2. 2x: https://www.snpedia.com/index.php/rs2180439(T;T)
3. \"increased risk\": https://www.snpedia.com/index.php/rs1385699(T;T)
4. 1.6x: https://www.snpedia.com/index.php/Rs1160312(A;G)
5. \"normal (higher) risk\": https://www.snpedia.com/index.php/Rs2003046(C;C)
6. \"normal (higher) risk\": https://www.snpedia.com/index.php/Rs8085664(C;C)
7. \"more likely\": https://www.snpedia.com/index.php/Rs6036025(G;G)
8. \"increased risk\": https://www.snpedia.com/index.php/Rs6625163(A;A)
9. \"more likely\": https://www.snpedia.com/index.php/Rs2223841(A;A)
10. \"more likely\": https://www.snpedia.com/index.php/Rs925391(C;C)

But #1 is a genoset (https://www.snpedia.com/index.php/Gs122/criteria) which already includes #8 and #4, so we shouldn't include those. This leaves us with 7x, 2x, and then several non-numeric \"more likelies\".

(I really don't know much about this, and could easily be misinterpreting things.)", "timestamp": "1544896805"}, {"author": "Jess", "source_link": "https://www.facebook.com/jefftk/posts/972501387302?comment_id=972518982042&reply_comment_id=972533687572", "anchor": "fb-972518982042_972533687572", "service": "fb", "text": "→ Thanks! Helpful.", "timestamp": "1544898683"}, {"author": "Danner", "source_link": "https://www.facebook.com/jefftk/posts/972501387302?comment_id=972562829172", "anchor": "fb-972562829172", "service": "fb", "text": "This was basically GenePeeks business. You really need a geneticist to speak to the level of observed mutations, just looking at the variants isn't good enough, because it's really protein production that matters, and it's the combination of mutations in both sets of your DNA that actually leads to the more subtle problems. Most companies avoid discussing illnesses because they don't have medical authority to do so. Everyone has busted parts, and if you just gave someone a list of things they might get or have deficiencies in, most people couldn't understand the risk profiles enough to have a sane discussion about it. Yes, Jeff Kaufman, I think you could handle discussing risk, so... talk to a genetic counselor?", "timestamp": "1544907460"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/972501387302?comment_id=972562829172&reply_comment_id=972563228372", "anchor": "fb-972562829172_972563228372", "service": "fb", "text": "→ Much more than most of medicine, though, this seems like something we should be able to automate", "timestamp": "1544907527"}, {"author": "Danner", "source_link": "https://www.facebook.com/jefftk/posts/972501387302?comment_id=972562829172&reply_comment_id=972563403022", "anchor": "fb-972562829172_972563403022", "service": "fb", "text": "→ GenePeeks was doing that, but ... the world isn't ready. 4 pivots, no takers. I know it has value and what we were doing is part of the future, but the timing and location just wasn't right yet.

Check out Lee Silver's work though.", "timestamp": "1544907627"}, {"author": "Daniel", "source_link": "https://www.facebook.com/jefftk/posts/972501387302?comment_id=972572649492", "anchor": "fb-972572649492", "service": "fb", "text": "When a gene mutation gives you 2x risk of getting a disease, does that mean twice the risk of an average person (whose genetic status is unknown)? Or twice the risk of a person who is known not to have the mutation? These two things could be significantly different, depending on the prevalence of such mutations.", "timestamp": "1544913684"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/972501387302?comment_id=972572649492&reply_comment_id=972580239282", "anchor": "fb-972572649492_972580239282", "service": "fb", "text": "→ Daniel I think the latter? But I don't know.", "timestamp": "1544918441"}, {"author": "Bucky", "source_link": "https://www.lesswrong.com/posts/XAwY5czzfkNNQohDJ#bJGCSdNvMQjeo5Lzy", "anchor": "lw-bJGCSdNvMQjeo5Lzy", "service": "lw", "text": "Thanks for writing this, it's not something I'd looked at before but I read some of the Promethease sample reports because you got me interested.

There does seem to be some weird normalisation going on when calculating magnitude. For instance this gene gives a score of 0 for (C;C) and bad magnitude of 2.7 and 3.1 for (C;G) and (G;G) respectively. So if you have (C;C) and you filter by magnitude 2 you miss out on the fact that you have an advantageous genotype.

This isn't a problem if (C;C) is extremely common but actually it's no more common than (C;G) (except for people of African descent), so the act of filtering prevents you from realising that you missed a 50:50ish chance of getting a disadvantageous genotype.

So probably to work this out properly you can't filter by magnitude and you'd have to open up every genoptye details to check for what you've avoided getting hit by. You could only really work out how well you've done compared to other people where the data includes frequency so you could see just how lucky/unlucky you got for a particular gene.

Not all of the genotypes have this issue - for instance this gene seems to be more sensibly normalised. If they were all done like this then I'd be much happier with the system.", "timestamp": 1545047726}, {"author": "Andy", "source_link": "https://www.facebook.com/jefftk/posts/972501387302?comment_id=973417705992", "anchor": "fb-973417705992", "service": "fb", "text": "On a side note, idk why people aren\u2019t more worried about compounding deleterious mutation accumulation, since it would logically increase every generation...", "timestamp": "1545418221"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/972501387302?comment_id=973417705992&reply_comment_id=973569531732", "anchor": "fb-973417705992_973569531732", "service": "fb", "text": "→ I'm not very worried about this because I think we're not very many generations away from being able to fix these.", "timestamp": "1545502388"}]}