Fine mapping with deterministic approximation of posterior Wen 2016

[gaow]

@stephens999 I've been trying to figure out how to take advantage of DAP in m&m:

https://dx.doi.org/10.1016%2Fj.ajhg.2016.03.029

However it is not immediately clear to me how this can be applied to multivariate setting of the m&m case. Also my understanding is that we want posterior of beta (J effects on R conditions) as a result of fine mapping, but Wen 2016 gives posterior of configurations. The DAP idea exploits the fact that only very few of all possible configurations actually are relevant. I am not sure how this can be directly used in m&m.

On the other hand, once we have hyperparameter for the prior we are only left with beta (J x R) and MCMC for beta posterior should be a lot simpler. Would you still see it not feasible to just run MCMC? Or, by referring to the DAP paper you are asking us to think about ways similar in idea to DAP in order to speed up the computations?

[stephens999]

I think mcmc may be feasible. Matthew

…

On May 1, 2017 5:51 PM, "gaow" ***@***.***> wrote: @stephens999 <https://github.com/stephens999> I've been trying to figure out how to take advantage of DAP in m&m: https://dx.doi.org/10.1016%2Fj.ajhg.2016.03.029 However it is not immediately clear to me how this can be applied to multivariate setting of the m&m case. Also my understanding is that we want posterior of beta (J effects on R conditions) as a result of fine mapping, but Wen 2016 gives posterior of configurations. The DAP idea exploits the fact that only very few of all possible configurations actually are relevant. I am not sure how this can be directly used in m&m. On the other hand, once we have hyperparameter for the prior we are only left with beta (J x R) and MCMC for beta posterior should be a lot simpler. Would you still see it not feasible to just run MCMC? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#9>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABt4xXr5R4V_Aad51SdLbeoZDjmkiIzuks5r1mHxgaJpZM4NNhr3> .