News for hackers. Stuff that matters.
Re:Association between polygenic risk scores for ADHD and co... by AlexCoventry on Tuesday February 21, 2017 @05:49PM (#13700913)
Has GWAS become a reliable means of identifying causal variants? How have the methods evolved over the last five years? What are the big success stories?
Re:Association between polygenic risk scores for ADHD and co... by a_bonobo on Tuesday February 21, 2017 @06:58PM (#13701481)
I work a bit in the field - GWAS is still not identifying causal variants, only linked variants.

There are tons and tons and tons of publications that keep on getting published on running similar datasets with the same methods (PCA, regression or MLM, done) and then reporting the associated SNPs. In humans these papers are getting huge with massive datasets, like this one: http://biorxiv.org/content/early/2015/05/27/019885

There was also this interesting paper which relied on a very specific configuration of a human population, you cannot do that with for example Brits: http://biorxiv.org/content/early/2015/11/12/031518

By its very nature GWAS cannot prove a causal relationship, I'm now hoping that CRISPR-Cas9 can provide for that - use GWAS to find candidate SNPs, use CRISPR to introduce those SNPs in an unaffected population, measure phenotype changes. Of course with sequencing getting so cheap you can start with SNP calling and get a much more comprehensive picture by looking at large scale insertions and deletions by just looking at the whole genome, like this paper for 10,000 (!!!) humans: http://biorxiv.org/content/early/2016/07/01/061663

Re:Association between polygenic risk scores for ADHD and co... by AlexCoventry on Tuesday February 21, 2017 @10:48PM (#13702841)
Thanks. Are the GWAS's the primary justification for the sequencing efforts generating the data they're analyzing?
Re:Association between polygenic risk scores for ADHD and co... by a_bonobo on Wednesday February 22, 2017 @01:30AM (#13703528)
I'd say the primary motivation for the sequencing efforts is that we're hitting the limits at what we can detect using SNPs. We now know of many inversions, deletions, insertions which are linked to phenotypes, and often you don't have a SNP positioned close enough to detect a signal. It's just easier and more comprehensive to align the genomic reads directly to detect these rearrangements than it is to look for a null signal in the SNP dataset.
Re:Association between polygenic risk scores for ADHD and co... by AlexCoventry on Wednesday February 22, 2017 @06:29PM (#13710262)
What do you do with those associations, once you've found them? Are they helping to elucidate the biology of the phenotype?
Re:Association between polygenic risk scores for ADHD and co... by a_bonobo on Sunday February 26, 2017 @10:09PM (#13741562)
There are many things you can do with the association, here's a non-exhaustive list:

- do the SNPs cause changes in a gene of interest? are the changes synonymous (cause base change, but the protein amino acid is identical since several codons code for the same amino acid, but this can cause problems in mRNA folding)? Are the changes non-synonymous (cause base change, and a different amino acid)? How strong are the non-synonymous changes? It could be that you get a shorter protein because a SNP introduced an early STOP codon, or changes a hydrophile to a hydrophobe amino acid, etc. From there you can check whether this change influences the phenotype by introducing this changed protein into another individual to see whether the phenotype is reproduced, this rarely works in more complex species as you don't have a single gene -> single phenotype cause

- you can use the ratio of synonymous/non-synonymous SNPs to approximate selection: genes under positive selection show more non-synonymous changes, these are usually genes involved in resistance who have to keep up with fungal/viral invaders, or venom genes, etc., genes under negative selection show more synonymous changes, these are usually housekeeping genes or other single copy genes that are important

- are the SNPs closely upstream or downstream from a gene of interest? It's still a bit hard to find binding sites for transcription factors (esp. outside of model species), so the SNP could be in a binding site where the whole transcription apparatus binds - do we see different transcription level for this gene in other individuals without the SNP?

- after DNA folding, is the SNP close to an interesting gene? Only with new HiC data can we test this, SNPs that were >10,000 base pairs away are after folding right next to the gene

- is the SNP in Linkage Disequilibrium with something interesting? In other words, is the SNP inherited together with something interesting? This could point towards selection