Question

clusterProfiler - qvalue adjustment method in enricher()

0

Entering edit mode

Jane Merlevede ▴ 90

@jane-merlevede-5019

Last seen 5.5 years ago

Hello,

Does someone know what is the adjustment method used when referred to qvalueCutoff argument of enricher(GO/DO)() function?

I tried pAdjustMethod = "holm", "hochberg", "hommel", "bonferroni", "BH", "BY", "fdr", but in every case, p.adjust and qvalue are different.

Does the qvalueCutoff apply to the qvalue and not to the p.adjust?

My qvalues and p.adjust are too close to help answering this question.

Thank you

clusterProfiler • 3.2k views

ADD COMMENT • link updated 7.4 years ago by Guangchuang Yu ★ 1.2k • written 7.4 years ago by Jane Merlevede ▴ 90

score 0 · Answer 1 · 2016-11-09

0

Entering edit mode

Guangchuang Yu ★ 1.2k

@guangchuang-yu-5419

Last seen 27 days ago

China/Guangzhou/Southern Medical Univer…

qvalueCutoff is only apply to qvalue and pvalueCutoff apply to pvalue and p.adjust.

p.adjust and qvalue are calculated independently using different methods. p.adjust will change if you use different pAdjustMethod but qvalue will not.

> library(DOSE)

> data(geneList)
> de =names(geneList)[1:100]
> x = enrichDO(de)

> y= enrichDO(de, pAdjustMethod="bonferroni")

> as.data.frame(x)[as.data.frame(y)[,'ID'], 'qvalue'] == as.
> xx = as.data.frame(x)
> yy = as.data.frame(y)
> all(xx[yy[, 'ID'], 'qvalue'] == yy[,'qvalue'])
[1] TRUE

ADD COMMENT • link 7.4 years ago Guangchuang Yu ★ 1.2k

0

Entering edit mode

Thank you for the answer.

Concerning my first question method used when referred to qvalueCutoff argument, since you replied that p.adjust and qvalue are calculated independently using different methods, I searched furthered on the web and found a discussion that brought me to the possible answer. It may be interesting for other users: https://www.biostars.org/p/150108/

If analyzing an ontology with a hierarchical relationships, these correction factor (FDR/FWER) will be skewed because genes in a child ontology term will also be annotated to its parent and thus are not true independent measurements. A correction factor like Elim or Weight (Alexa et al 2006) take these dependencies into account when performing these correction (andrew).

The qvalue is the result from another type of correction, different from FDR and FWER that takes into account the recurrence of genes in different terms. So, I assume the qvalue is the result from Weight method mentionned in this paper. Am I right?

ADD REPLY • link 7.4 years ago Jane Merlevede ▴ 90