hypRcoloc with 1500 phenotypes across 200 loci

Question

hypRcoloc with 1500 phenotypes across 200 loci

Closed this issue 4 years ago · 2 comments

More of a question than an issue, but here we go:

I have run hypRcoloc on a dataset with ~1500 phenotypes measured on the same individuals. There are ~200 loci where at least 1 phenotype has genome-wide significant P-value < 5e-8. To make computations easier (/not crash my PC), for each loci, I extracted all phenotype with at least a suggestive P-value < 1e-6 and ran hypRcoloc with default parameters.

Results look pretty good, but I'm wondering about the priors: In the preprint, you run hyprcoloc in predefined LD-blocks. In my case, I run hypRcoloc at loci where I know at least one phenotype has significant signal.

Should I change prior.1 or prior.2 to something other than defaults in this case?

Answer 1 · 2020-12-09T11:58:38.000Z

Hi Malte,

I think the default priors should be okay in your situation. @cnfoley anything to add here?

Best wishes,

James

Answer 2 · 2021-03-12T09:25:56.000Z

Hi Malte, With so many traits and regions, I think it would be sensible to assess sensitivity to the choice of priors. What I would do is: 1. Run default analysis in each region 2. (Within a region) For each cluster of colocalized traits identified - run a separate analysis assessing sensitivity to the default prior using the "sensitivity.plot". Section 3 in the vignette covers this. If there are too many clusters for (2) to be pragmatically feasible let me know and I'll send you a function which avoids plotting. For this, you will have to input a 'decision rule' - deciding how much sensitivity to the prior (i.e. how sensitive the custer structure is to the prior) that you are willing to accept. Good luck. Chris

…

On Wed, Mar 10, 2021 at 8:40 PM James Staley ***@***.***> wrote: Hi Malte, I think the default priors should be okay in your situation. @cnfoley <https://github.com/cnfoley> anything to add here? Best wishes, James — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#14 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALJAQVMG5X2TVSM2FRV4F5DTC7KK7ANCNFSM4T7SJJPQ> .