search

bulik / ldsc

LD Score Regression (LDSC)
GNU General Public License v3.0
639 stars 342 forks source link

Z-scores New Categories #92

Open sbguarch opened 6 years ago

sbguarch commented 6 years ago

Hi,

Just to be sure.

I calculated the functional enrichment for my own 12 functional categories (I followed the tutorial that you gave for cell-type specific annotations). I took the final Z-scores for each annotation, well that I transformed to p-values, which is the result that I finally plotted. (I think that this is similar to the figure for cell-type specific enrichments from you main paper Fiuncane HK et al 2015).

Do you think that this is an appropriate way to compare functional enrichment between my own categories?

Thanks for your feedback.

Thanks.

rkwalters commented 6 years ago

Hi, That generally sounds reasonable. The two things I’d watch for are: 1) whether or not it makes sense to be conditioning on the “baseline” categories that are used with the cell-type specific analyses (e.g. if you’re annotations are partially redundant with some of those baseline annotations). 2) whether you’re interested in the test for marginal enrichment or the test for enrichment conditional on the other annotations (i.e. the test of the coefficient) Cheers, Raymond

On Oct 25, 2017, at 4:46 AM, sbguarch notifications@github.com wrote:

Hi,

Just to be sure.

I calculated the functional enrichment for my own 12 functional categories (I followed the tutorial that you gave for cell-type specific annotations). I took the final Z-scores for each annotation, well that I transformed to p-values, which is the result that I finally plotted. (I think that this is similar to the figure for cell-type specific enrichments from you main paper Fiuncane HK et al 2015).

Do you think that this is an appropriate way to compare functional enrichment between my own categories?

Thanks for your feedback.

Thanks.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/bulik/ldsc/issues/92, or mute the thread https://github.com/notifications/unsubscribe-auth/AILEvU_c8oZwxdgIkJ2NgA9l2WH_4nKJks5svvVPgaJpZM4QFo8y.

sbguarch commented 6 years ago

Hi , Thanks for you answers! Regarding your points. 1) I thought that it would be worth correcting for the baseline model because my annotations are generated based on our specific epigenomic datasets, and moreover some categories may overlap at some extent between them. 2) the aim is to test the contribution of my categories. Taking into account this, I thought that it made sense focusing on the category-specific per snp heritability (test of the coefficient, the z-scores) instead of looking only at the enrichment (prop h2/prop snps). Did I get that right?

rkwalters commented 6 years ago

Yep, that sounds right. Best of luck with the analysis!

Cheers, Raymond

On Oct 26, 2017, at 5:53 PM, sbguarch notifications@github.com wrote:

Hi , Thanks for you answers! Regarding your points.

I thought that it would be worth correcting for the baseline model because my annotations are generated based on our specific epigenomic datasets, and moreover some categories may overlap at some extent between them. the aim is to test the contribution of my categories. Taking into account this, I thought that it made sense focusing on the category-specific per snp heritability (test of the coefficient, the z-scores) instead of looking only at the enrichment (prop h2/prop snps). Did I get that right? — You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/bulik/ldsc/issues/92#issuecomment-339811904, or mute the thread https://github.com/notifications/unsubscribe-auth/AILEvZ2X2CArfSqy31aSA6VL-YtlR_bQks5swP89gaJpZM4QFo8y.