eQTL maximum P-value (≤) SNP2GENE function

[chris....@c4xdiscovery.com]

Hi there,

New to using FUMA, been looking over the FUMA paper and tutorial site. But getting confused on the cost and benefit of using FDR or nominal pvalues for eQTL.

We are looking at the value of using either the default setting FDR ≤ 0.05  compared to changing nominal p-value as a way to filter SNP to gene mapping. 

I have seen that the eQTL databases are compiled from various sources (n=8), mainly having the FDR scores (MuTHER has none), does this mean using the nominal p-values is ill advised as we loose some databases and the way some of the p-values are calculated seem to differ between sources.

This is to help reduce the number of genes being associated with the lead SNPs and their corresponding LDs.

If FDR is the preferred filtering method in eQTL, is their any suggestions to help reduce the SNP2GENE association to have highly correlated genes?

Kindest regards,

Chris

  

[Kyoko Watanabe]

Hi Chris,

Thank you for your interest in FUMA.
You are right, P-value and even how FDR was computed differ across data resources so there is no "golden standard" way to filter eQTLs based on a single value.
If you want to play around the P-value or FDR threshold, I would suggest you to run a FUMA with P-value threshold let's say 1e-5 and you can see how other filtering value affect by downloading eqtls.txt file.

>This is to help reduce the number of genes being associated with the lead SNPs and their corresponding LDs. If FDR is the preferred filtering method in eQTL, is their any suggestions to help reduce the SNP2GENE association to have highly correlated genes?

For this, you need to perform colocalization. Currently FUMA just annotate eQTLs as long as eQTL exist within the certain LD of independent significant SNPs. But colocalization help you to further prioritize which gene is the most likely affected by eQTL. There are several software available, for example you can try SMR (http://cnsgenomics.com/software/smr/#Overview).

Best,
Kyoko