molgenis/systemsgenetics

How to select significant cell type eQTL

zhangpicb opened this issue · 6 comments

zhangpicb commented

Dear @npklein

Thanks for your great tool!

I want to use Decon-qtl to find cell-type eQTL with cibersort results,But I find no significant p-value after B-H correction. Most p-value in results are 1,like README

systemsgenetics/Decon2/README.md

Lines 151 to 153 in a7fc9e2

| |Granulocytes_pvalue |B cells (CD19+)_pvalue |CD4+ T cells_pvalue |CD8+ T cells_pvalue |NK cells (CD3- CD56+)_pvalue |Monocytes (CD14+)_pvalue |Beta1_Granulocytes |Beta2_B cells (CD19+) |Beta3_CD4+ T cells |Beta4_CD8+ T cells |Beta5_NK cells (CD3- CD56+) |Beta6_Monocytes (CD14+) |Beta7_Granulocytes:GT |Beta8_B cells (CD19+):GT |Beta9_CD4+ T cells:GT |Beta10_CD8+ T cells:GT |Beta11_NK cells (CD3- CD56+):GT |Beta12_Monocytes (CD14+):GT
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---
ENSG00000103855\_snp_244 |0.8754 |1.0 |0.9855 |1.0 |0.8801 |1.0 |0.0 |0.0 |0.0 |0.0 |0.1954 |0.0 |-0.0083 |0.0 |-0.0014 |0.0 |0.0581 |0.0

How to select significant cell type eQTL?

Thanks in advanced!

npklein commented

Dear @zhangpicb

How many samples did you use? The reason that they are all not significant in the README is because it is a toy example with a very small data set. To identify interactions at least a couple hundred samples are necesarry.

zhangpicb commented

Dear @npklein

Thanks for your quick reply!

The sample number is 93.

zhangpicb commented

I find the method in Decon-qtl is similar to GTEx cell type interaction eQTL .Becuase our lab own GTEx v8 data,without BLUPRINT data.So I replicate https://www.science.org/doi/10.1126/science.aaz8528 Fig5c , But the pvalue is 1.

npklein commented

Our method is quite a bit different from GTEx method. In our paper we show that we trade higher specificity for lower sensitivity, so that could be why it does no replicate.

Alternatively, can also be due to strict pvalue correction? For how many SNP-Gene pairs did you run it? We run it only for top SNPs, so if you ran for all snp gene combinations that could also be the issue

zhangpicb commented

Dear @npklein

Thanks for your quick reply!

I use all snp-gene combinations,around 1500 snp-gene pairs per gene.The SNPs are from gene tss±1Mb.
How to select top SNPs for each gene?

Thanks in advanced!

npklein commented

Okay that is not too many tests, in that case I think more samples are necessary