Integrating MetaXcan with eQTL Databases Built Using TensorQTL

[Agathe Bugnon]

Dear Community,

Thank you for developing MetaXcan and setting up this Google Group. We have already used the tool for several projects using the eQTL databases from PredictDB. Now, we aim to apply MetaXcan with our own eQTL databases, which have not been produced using an Elastic Net or MASHR model.

Specifically, we have generated cis- and trans-eQTL databases using the Broad Institute's TensorQTL tool. However, since the association method differs, we do not have "weight" values for each pair, nor "pred.perf.R2" or "pred.perf.pval" values for each gene. Based on the MASHR model, we understand that "pred.perf.R2" or "pred.perf.pval" values might not be essential ; and we've substituted the "weight" value with the "regression slope" value from the association measurements performed by TensorQTL for each gene/SNP pair, but we are uncertain if this substitution is appropriate.

If anyone has encountered similar challenges or explored MetaXcan with databases such as those generated by TensorQTL, we would greatly appreciate discussing the methodology to ensure the compatibility of our approach.

Thank you in advance for your time and assistance.

Best regards,
Agathe Bugnon.