Tempted to start Millennium harvest house themed channels :upsidedownface:

We decided it was too much effort to arrange delivery of soggy sandwiches and expired Diet Coke to all participants.

Hi all, just want to make you aware that @here is where you can have the ear of a good chunk of the psychiatric genetics/statistical genetic and behaviour genetic community. Its almost like being at a conference where you can als questions discuss papers you may want to understand better etc etc, I'd liberally make use of that opportunity!

This would be precisely the type of chat I would try to initiate in front of a drink, so I guess the #random channel is the one to go.

What is the null in BG, and why does it matter?

After @Benjamin Neale intro <#C02093NH9K3|intro-genomics-biometric-model>, I left wondering what the null is in the field. Following the introduction, if I remember correctly, the null is that heritability = 0.

But I feel that, sometimes, the null of someone is someone else’s alternative (e.g., Null heritability != 0; Turkheimer, 2000).

I would love to hear the opinion of other experts on how setting different nulls can have an impact, if any, on the field. For example, when conducting a study, should we assume that any trait will show some heritability, even if we have no information on the trait of interest? And if so, are only results that deviate from this expectation “surprising”? Asking for a friend ( 🥸 ) who got a review on the lines of what outlined above.

SAGE Journals

SAGE Journals: Your gateway to world-class research journals

Subscription and open access journals from SAGE Publishing, the world's leading independent academic publisher.

Original URL: https://journals.sagepub.com/doi/abs/10.1111/1467-8721.00084?casa_token=zpkQCJPc20IAAAAA:ffLrLwNBSar0YnLhOKvrOoCYX3SLhyiEBxz-j7Ev_1jugcRk6obCkBh5ebxEm7cu8Qd_IrLZNILx-Q

*Thread Reply:* Great question Giacomo - I think it's worthwhile to delineate between the test of the null and what we think the world is actually - Turkheimer's point is that twin studies find heritability

*Thread Reply:* I think the null should depend on the question at hand. If you’re looking at a genetic correlation, whether the null is 0 or 1 (or any other value) just depends on the scientific question. For example: If you’re testing the same two traits across 2 samples, your null is likely to be rg=1. If you’re testing two different traits, it is likely to be 0. Finally, if there’s a study out there that shows rg=.5 between 2 traits and you have the same 2 traits but measured in a different context (or whatever), your null may well be rg=.5.

*Thread Reply:* and the null model shouldn't interfere with parameter estimation as a goal - in Matt's example, the standard error on the estimate of the rg matters a lot - if it is an rg of 0.5 with SE of +/- 0.2 then we'd have a z of ~2.5 that it is not 0 and a z of ~2.5 that it is not 1 so reasonable evidence that it is neither 0 nor 1, but not much that - contrast that with the circumstance of a SE of 0.02 - where we would put the likely r_g v. close to 0.5.

*Thread Reply:* One ramification of h2 != 0 as the null that Turkheimer points out is about making causal claims based purely on phenotypic comparisons between relatives (e.g., kids do good in school because their parents read to them). Making this kind of causal conclusion requires ruling genetics out as an explanation, or at least conditioning on relatedness.

*Thread Reply:* Going off Ben’s post, another point may be that we should be a lot less concerned about null hypothesis testing and a lot more concerned about effect size estimates & their SEs

@Jeff Lessem (he/him) Hi Jeff, how can I download the practice datasets to my own computer so I can try them at home? Thx

*Thread Reply:* The easiest way is to use an sftp or scp client, such as FileZilla, winscp, bitvise, or others you can search for. Then use workshop.colorado.edu as the host and your workshop username and password.

*Thread Reply:* You can also use Rstudio Server: login to https://workshop.colorado.edu/rstudio/, copy the files to your home directory using the Terminal tab (lower left), then select the files/folders in the Files tab (lower right), then click the "[gear] More" dropdown menu and select "Export..."

I'd like to save the scripts for future reference

For anyone who was confused by some of the shell commands, I find that this website is helpful for understanding what some of the flags mean: https://explainshell.com/

*Thread Reply:* I do like that this preserves some of the humour of the developers. ie less is the opposite of more :)

*Thread Reply:* https://learnxinyminutes.com/ is also a great one for checking how things work very quickly

Food for thought: There are SEVENTEEN questions about cereal preferences in the UKB. Yet no question about whether the person has ever been married. 🤔

*Thread Reply:* well... there is this (which is current not ever I realize: https://biobank.ndph.ox.ac.uk/showcase/field.cgi?id=6141 and this which is only last 2 years... https://biobank.ndph.ox.ac.uk/showcase/field.cgi?id=6145

*Thread Reply:* @Jared Balbona - check out the links from Michel above

*Thread Reply:* thanks Michel!

*Thread Reply:* Sure am glad we have such deep phenotyping on food preferences though 🙄

*Thread Reply:* @matthew keller and @Jared Balbona are you working on marriage/divorce? Would love to discuss further! Working with @Nathaniel Thomas @Sally Kuo and @Sarah Brislin (she/her) on nature of nurture models for alcohol use disorder via parental divorce and parental marital discord and have some interesting preliminary findings. Have also fit models using using the Balbona, Kim, and Keller models - some head-scratching results currently working through.

*Thread Reply:* Yes, would be good to catch up. We’re not working directly on marriage/divorce, but rather thought of a way to test mechanisms of assortative mating depending on comparing never married to married persons. Also, we’re starting to apply our models to the UKB and are preparing ourselves to be confused by real data as well. We’ve already caught a few issues just looking at the correlations of PGS haplotype scores between spouses

*Thread Reply:* 👍 we were planning to reach out to Jared in the near future, so will keep you all in the loop.

*Thread Reply:* I curse UKB phenotyping daily, but I have no illusion I would do better, or that any other sample will be as big/accessible/complete/related any time soon.

*Thread Reply:* This was in response to Matt not Jessica

*Thread Reply:* Hi, Jessica-- I’d also love to catch up sometime! The AUD/ divorce project is such a cool idea, so I’m excited to see the preliminary findings (and happy to help resolve any confusing ones as best as I can). Feel free to reach out anytime! 🙂

*Thread Reply:* Ask and you shall receive, Jared (Nate is drafting a message to you now...)

*Thread Reply:* @Michel Nivard - UKB has been a godsend!!!!! They’ve shown us the way in genetics. So I’m not bitching. OK, I guess I am a little bit. Their preoccupation with diet takes up about 20% of all the questions and is probably used in 0.2% of all the publications. Hope that person on the committee who pushed this is happy. But overall, UKB has been fantastic. Let’s see if NIH (AllOfUs) can ever approach this.

Anyone have any idea what is happening to the NIH AllOfUs program (the US’s attempt to do a bigger and better UKB). You never seem to hear about it anymore. Are they letting it die a slow death?

*Thread Reply:* Far from it. They are ramping up and expanding data collection efforts. They have a listserv. No genotype data available yet. Just phenotypic data. And have been hiring for quite a few positions.

*Thread Reply:* https://allofus.nih.gov/news-events-and-media/announcements/media-advisory-nihs-all-[…]nificant-progress-participant-diversity-and-research-underway

*Thread Reply:* I actually got a newsletter from them today, they just published a paper demonstrating a few positive covid cases in states prior to the state's first confirmed case using the blood samples they had collected in Jan-March 2020 https://academic.oup.com/cid/advance-article/doi/10.1093/cid/ciab519/6294073

Today's random trivia - to avoid spaces in the names of variables or files coders often name things with a lowercase on the first word and an Uppercase first letter of subsequent words. This is called camel case or camelCase.

*Thread Reply:* There's also snake_case, kebab-case, TRAIN-CASE, and SCREAMING_SNAKE_CASE

*Thread Reply:* nice 🙂

does anyone use R for building your CV? any favourite packages/scripts?

*Thread Reply:* I was planning to, but I have seen most solutions rely heavily on translating the code from RMarkdown to Latex (which is what I use for my CV atm). What I have seen that is most used around is the package vitae with the moderncv Latex theme, and rorcid to get automatic articles updates from your orcid page

*Thread Reply:* thanks for your take on this!

I would love to get some input on an issue I am having preprocessing some GWAS sumstats. I need to add MAF to a sumstats file that only contains chr:pos, A1, A2, b, and se. So far, I have downloaded the 1KG (EUR) reference panel (the same used in the original GWAS) and used plink to calculate MAF. But now I am left with the issue of merging those MAFs with my sumstats file. Here’s the rub: I am unsure how to check if the alleles are correctly aligned before doing that merge. Any thoughts?

Right now I have a .txt sumstats file and a .afreq file that I need to merge in.

*Thread Reply:* Hi Peter, do you have A1 frequency on your summary statistics file?

*Thread Reply:* Do you need MAF added to this OR do you need the Freq of the effect alleles - these are 2 different things and the way to do this differs depending on what you need. If you only need MAF you don't need to worry about which allele is the Minor Allele (ie A1 or A2)

*Thread Reply:* Nope. 😕 I only have the following columns: chr:pos, A1, A2, b, and se.

*Thread Reply:* @Sarah Medland (she/her) Yup, I misspoke (misswrote?), I need the effect allele frequency.

*Thread Reply:* And this is date you've gotten from somewhere or is this you're own GWAS where you also have access to the individual level data?

*Thread Reply:* @Gunn-Helen Moen I’m using publicly available stats.

*Thread Reply:* I do not have access to the individual level data

*Thread Reply:* I think my first course of action in that case would be to contact the author of the paper that these summary statistics come from and ask them if it is at all possible to get information regarding allele frequency. Often authors will be happy to share more data if they can. And that way you'll be certain that the frequency you're using are correct

*Thread Reply:* That was my first thought as well. Unfortunately, I have reached out about this issue and the point of contact said they could not provide the additional data. 😕 Now I’m looking for a technical solution.

*Thread Reply:* EDITED - If you need frequency of the effect allele & the GWAS was not run in plink. First drop all ambiguous SNPs. Then you need to get freq of A1 not the MAF. Once you have this you can merge on SNP-ID and A1. Extract those that didn't match. Assuming you want to keep A1 as the effect allele across the file that has the GWAS results, calculate allele freq of A2 (1 - freqA1) and merge this on SNP-ID and A2. Any SNPs that are still unmatched can be manually checked for annotation errors or dropped. (I always code this as needed as the file formats are never constant but these are my steps).

*Thread Reply:* Thank you, Sarah! This has been a real headache, I appreciate your help!

*Thread Reply:* Regarding some authors' reluctance to share allele frequency information: there is a privacy angle that if you have access to someone's genotype and FRQ from a study's summary statistics, you can probabilistically infer whether that individual was a participant in the study. This has become less of a concern as sample sizes have grown, however.

*Thread Reply:* I think it's also unclear how well the Homer et al type methods under contemporary meta analysis conditions - large N, 20+ cohorts & complex missing data structures due to cohort level QC

*Thread Reply:* Does Plink automatically assign the minor allele to A1? I ask because I downloaded the 1KG reference panel from the Plink2 website (link), and of course it comes in Plink2 format. If I am interested in the effect allele that all of the studies from the GWAS imputed to (and not necessarily the minor allele) do I need to download a non-plink version of the 1KG panel? In order to avoid that auto-assignment of alleles?

*Thread Reply:* Plink2 no longer auto assigns the minor allele to A1, as Plink1 did by default.

… fake or real?

Image from iOS

*Thread Reply:* Lol - real evidently: https://untappd.com/b/new-image-brewing-genetic-nurture/3463391

Untappd

Genetic Nurture - New Image Brewing - Untappd

Genetic Nurture by New Image Brewing is a IPA - New England which has a rating of 4.3 out of 5, with 992 ratings and reviews on Untappd.

Original URL: https://untappd.com/b/new-image-brewing-genetic-nurture/3463391

*Thread Reply:* WAHH??

*Thread Reply:* someone should send Augustine Kong a case...

*Thread Reply:* I don’t drink beer (and especially IPA) but because of the label, I need to try it!!!

*Thread Reply:* New Image Brewing is apparently in Arvada, so Boulder folks could actually go get some!