I just uploaded the final set of notes on genomic prediction. As you’ll see, there are some things you’ll need to be careful of if you want to use genomic prediction in your own work. Although I don’t mention this explicitly in the notes, the implication of the last caveat about genomic prediction is that GWAS may not be very useful in understanding the genetic basis of variation in a complicated trait beyond the population in which the analysis was done. A further leap is that even a traditional quantitative genetic analysis may not extrapolate well to other populations, even if the environmental conditions are similar.
In short, quantitative genetics provides some very powerful tools, but powerful tool can also cause a lot of damage. Be careful when you wield this one.
Congratulations! You still have a couple of lectures to survive, but you’ve almost joined the survivors of EEB 5348. I’ve now posted the last lab exercise for the semester. It builds on the locus-by-locus GWAS in last week’s lab and extends it to genomic prediction. We could easily spend several weeks, probably even an entire semester, exploring all of the ins and outs of GWAS and genomic prediction, but this is an overview course. My goal is to give you enough of an introduction to approaches like these that you not only know they exist but that you also can figure out how to use them on your own if it seems appropriate for your research.
If you downloaded the notes on genomic prediction before noon on Sunday, I suggest that you download them again. I’ve modified them a bit to reflect the new R notebook I put together on Saturday to illustrate the simple example that is discussed in the notes. I also suggest that you download the R notebook so that you can explore different simulated data sets and see how they perform.
I am working on the final lab exercise, and I’ve finished running nearly all of the analyses I want to run before typing it up and making it available. Since I have some non-teaching related things to attend to this afternoon, there’s a decent chance that the lab won’t be posted until some time on Monday, but I’ll do my best to have it posted before lecture on Tuesday.
It took me a bit longer to make sure the data and associated scripts were performing the way I wanted them to, but the lab exercise for week 13 is now posted. You’ll be running a genome-wide association analysis on a subset of data collected in an analysis of gypsy moths (Lymantria dispar). This is only an example. You can look at the Stan code if you want. If you do, you’ll see that it’s just the linear regression that we’ll discuss in class on Thursday with some magic added to take care of the relatedness among individuals. If you were doing a real GWAS, I wouldn’t recommend using this script. I’d recommend using GEMMA or something like it that was built by experts in GWAS specifically for the purpose of GWAS. There are a lot of things to be careful about that aren’t built into these simple scripts, but they illustrate the principle so that you can get a sense of how GWAS might be used in your own research.
As usual, you can find a link to the lab from the Lecture Schedule page (or you can in a few minutes when I get it posted) or from the link here.
I found the (rather obscure) error in my R Shiny app illustrating resemblance between relatives for a quantitative trait. You can find a link to the app either on the Lecture Detail page or below.
Note: It’s the same link before. The code inside the link has been changed, but not the link itself.
We found an error in the equation describing the covariance in half-sibs as I was lecturing on Thursday. I just verified that the error is also in the accompanying notes – not surprising since the lecture slides are directly extracted from the lecture notes. I’ve now corrected the notes and uploaded new versions. You can find them either from the Lecture Detail page or from the links below. The correction will also appear in the book version of the notes that I’ll compile at the end of the semester.
- Resemblance among relatives (HTML)(PDF)
It took me longer than it should have, but I finally added a link to the app I used in Tuesday’s lecture to illustrate partitioning of the genetic variance into additive and dominance components. You’ll find the link below and in the lecture detail page for Tuesday and for Thursday. In the lecture detail page for Thursday you’ll also find a link to an app illustrating the resemblance between parents and offspring. As noted there, I discovered that there’s a problem in simulations involving non-additive allelic effects. I plan to investigate and fix the problem over Thanksgiving break. I’ll post a note here when I’ve fixed it.
Lab 12 has now been posted to the course website. As usual, you’ll find a link below and from the Lab Schedule page. I spent nearly my whole weekend working out a few details that I realized I didn’t quite understand to make sure that this exercise makes sense. Unfortunately, that means that I still haven’t done any grading. I almost certainly won’t have time this week, so you can guess what I’ll be doing over Thanksgiving break. I apologize for falling so far behind in grading. The good news is that I’ve taken a quick look at a few of the exercises that have been turned in, and if that sample is representative, everyone has a good handle on everything we’re doing.
See you on Tuesday.
As a reminder, the lab exercise that is due on Friday, 11 December is the last exercise for the course and there will not be a final course evaluation. Once you’ve turned in the exercise for Lab 14, you’ve completed everything I expect. I hope when you turn it in you’ll be able to say as one of my former students did, “This was a good thing to have done.”
I mentioned in passing a few weeks ago that Sohini Ramchandaran will be presenting a seminar in EEB this fall. As a matter of fact, she’ll be presenting her seminar this Thursday at 3:30pm via Zoom. The timing of her seminar is perfect. (You might almost think I planned it this way.) She’ll be discussing how to make inferences about the genetic basis of complex traits, and you’ll have a chance to see some very modern approaches to quantitative genetics. It provides a preview of where we’ll be heading from Thursday morning through the end of the semester.
By the way, Sohini and I are academic siblings, although she’s much younger (and more accomplished) than I am. Our major advisor was Marcus Feldman.
I’ve just posted Project 3 on the course website. As with Project 2, you won’t need to analyze any data or run any simulations. Instead, you’ll need to read a paper, think about the techniques the authors used to detect natural selection (which are different from those we discussed in lecture), and answer some questions about the analysis and its implications. You can find a link to the project on the Lab Schedule page or you can follow the link below.
In other course-related news, I’ve included a YouTube video on the lecture detail page for Tuesday’s lecture. It provides a brief overview of sparg, the approach to inferring the spatial location of ancestors and the dispersal history from individual-level data. We’ll discuss sparg on Tuesday, but you may find it helpful to review the video before then.
McVicker, G., D. Gordon, and P. Green. 2009. Widespread genomic signatures of natural selection in hominid evolution. PLoS Genetics https://doi.org/10.1371/journal.pgen.1000471
IMPORTANT NOTE: The link to the paper was working Saturday afternoon, but it seems to be broken now (2:45pm, 7 November). The error message says this is “a likely temporary condition.” I’ll keep an eye on it. If it isn’t fixed soon, we’ll have to regroup. I also tried to get to some other papers on the PLoS Genetics website, and it appears to be affecting the whole site. The error message mentions a server configuration issue.
Update 8:05am, 8 November: I don’t know when the PLoS Genetics site came back up, but it’s up now. If you tried to get to the paper before and couldn’t, you should be able to get to it now. I’ve also downloaded a PDF that I can share if we run into trouble again.