Hi everyone.

I performed RNA-seq experiment on two independent groups for Group1 and Group2. number of Group1 samples is 15 and number of Group2 samples is 4. and number of genes is 20000.

And I am doing rna-seq data analysis using mt.maxT function of multtest R package (Permutation 5000).

this is summary data

summary(PP$index)

Min. 1st Qu. Median Mean 3rd Qu. Max.

1 5126 10251 10251 15376 20501

summary(PP$teststat)

Min. 1st Qu. Median Mean 3rd Qu. Max. NA's

-5.5036 -1.1150 -0.0513 -0.0511 0.9765 8.0654 878

summary(PP$rawp)

Min. 1st Qu. Median Mean 3rd Qu. Max. NA's

0.0004 0.1404 0.3884 0.4334 0.6998 1.0000 878

summary(PP$adjp)

Min. 1st Qu. Median Mean 3rd Qu. Max. NA's

0.0580 1.0000 1.0000 0.9992 1.0000 1.0000 878

summary(PP$fdr)

Min. 1st Qu. Median Mean 3rd Qu. Max. NA's

0.2102 0.5610 0.7761 0.7387 0.9330 1.0000 878

my smallest fdr p-value among 20000 genes is 0.21 and unadjusted p-value among 20000 genes is 0.0004.

In this situation, can I set cutoff using raw p value (unadjusted p value)?

The most common practice is to apply the Benjamini-Hochberg method to adjust the p-values for multiple testing, after differential analysis with edgeR / DESeq2 / limma-voom. Do you have a compelling reason to use a different correction procedure? And are you using the standard tools to perform the differential gene expression analysis?

Thanks h.mon.

I can't use other tools (edgeR or DESeq2). Because My data is quartile normalized by TCGA pipeline (no TCGA data). Additionally, I have only rsem quartile normalized data. :(