Understand about the output of interactionQTL

[broadinstitute]

Understand about the output of interactionQTL

[shengxin321]

But the README file has very few things about the interactionQTL.

[francois-a]

Hi,

Yes, that's the correct interpretation of these terms. Please see the README for additional details. The default cis-window is indeed TSS ± 1Mb; you can see the default options when running python3 -m tensorqtl --help.

[jackhump]

Dear Francois,

Could you give a little more detail on how you recommend correcting interaction QTLs for multiple testing? I've generated cell-type interaction QTL results using some deconvolution estimates (from MuSiC) that I residualised against batch, age, sex, and technical covariates in my RNA-seq data. However, every gene's P-value for the pval_gi term in the _top_assoc.txt.gz file seems to be significant after either Benjamini-Hochberg or qvalue correction (adjusted P < 0.05). Apologies if I'm misunderstanding something here.

[francois-a]

Hi,

Sure -- please see the added information in the README. I just realized that eigenMT was not run by default when running tensorQTL from the CLI. The latest commit fixes that. Otherwise, if importing tensorQTL as module, make sure to use the option run_eigenmt=True in map_nominal. You should then see the proper adjusted p-values in the column pval_adj_bh.