joint covariance for 49 tissues in GTEx8

[stat-yyang]

Hi,
I am implementing the TWAS test on your weight files. I successfully ran a single tissue analysis but not sure how to combine the test results for all tissues.
In UTMOST, they provide joint covariance across SNPs used for all genes, and then GBJ test is used to compute a single p-value for a pair of gene-phenotype.
Could you also provide joint covariance for the multi-tissue test and enable joint test across all tissues?
Thanks!

[zdangm]

In JTI, we didn't include the step combining the significance across all the tissues. (Instead, we included the pleiotropy control as a follow-up). However, one can do it using JTI model for sure. The genetically determined gene expression can be estimated using genomic data from reference datasets (e.g., 1000 genome project) and the prediction model (.db file) for each tissue. Both of them are publicly available. Then the covariance can be estimated among tissue-tissue pairs. Please let me if you have any difficulties or have any other questions.

[forget999]

Thanks for your great approach! I would like to know the more method details for estimating the covariance between tissue-tissue pairs that your mentioned. Thanks!

In JTI, we didn't include the step combining the significance across all the tissues. (Instead, we included the pleiotropy control as a follow-up). However, one can do it using JTI model for sure. The genetically determined gene expression can be estimated using genomic data from reference datasets (e.g., 1000 genome project) and the prediction model (.db file) for each tissue. Both of them are publicly available. Then the covariance can be estimated among tissue-tissue pairs. Please let me if you have any difficulties or have any other questions.