significant p value

[sumageb]

Hi, I am currently using matrix-eQTL for eQTL analysis of my wheat dataset. Is there a way to determine the best threshold for identifying cis and trans eQTLs? For example, how can I choose the value for pvOutputThreshold_cis and pvOutputThreshold_trans? I used 10-5 in both cases, but it provided many cis and trans eqtl, which may not be true. Also, I used the min.pv.by.genesnp = =T option, it provided me the p value number for each gene and snp, but I am confused about which one I should choose for the threshold. maybe I am missing something. Please guide me in this regard. Thank you.

[andreyshabalin]

Hi Suma Bala Dabi,

The p-value threshold can be based on the number of local and distant tests. Those numbers are reported by MatrixEQTL in the returned me object.

Andrey

[sumageb]

Hi Andrey, Thank you so much for your reply. I checked the number of tests for cis and trans, matrix_eqtl$cis$ntests 563866 matrix_eqtl$trans$ntests 2871359314 I am sorry for the naive question, but how do I decide the p-value from this value? Should I divide it by the number of samples or something else?

[andreyshabalin]

A good p-value threshold to start with would be pvOutputThreshold_cis = 1/matrix_eqtl$cis$ntests and pvOutputThreshold_trans = 1/matrix_eqtl$trans$ntests. It's also very helpful to inspect the QQ-plots for local and distant tests.

[sumageb]

Thank you so much for your help. I am closing this issue.