Regarding the use of mashr to evaluate the similarity of effect sizes under different conditions

[yrjia1015]

I am working with eQTL data from four distinct conditions and would like to assess how similar the effect sizes are for each gene across these conditions, specifically regarding both sign and magnitude. My goal is to calculate the proportion of signals that share the same sign and similar magnitude across conditions.

To achieve this, I am considering using posterior samples from mashr. However, I am unsure whether posterior sampling is necessary for this analysis, or if there is a more appropriate method for quantifying effect size similarity across conditions.

Additionally, I would like to understand if this approach differs from the traditional get_pairwise_sharing function, which calculates effect size sharing based on posterior means without utilizing posterior samples. Could you clarify if posterior sampling is essential for my analysis, or if the standard get_pairwise_sharing function would suffice?

[pcarbo]

@yrjia1015 I think the "mash sampling" vignette answers your last question:

We can also get the proportion of significant effects have the same sign and similar magnitude for each
pair of conditions. For each pair of conditions, first identify the effects that are significant in at least
one of the two conditions. Then compute the probability of sharing in sign and magnitude… it is computed
out of significant signals only.

If you would like to estimate sharing from your entire data set — and not just the significant results — then you would need to use get_pairwise_sharing_from_samples(). So it depends on what your analysis aim is.

Hope that helps.