why two packages have not similar FDR ???
1
0
Entering edit mode
8.6 years ago
unique379 ▴ 120

Hello folks, I am wondering why deseq2 and edgeR results are not agreed to each in terms of FDR while both calculated almost same LogFC and Nominal/Raw p-values ?? It seems DESeq2 has more stringent view to compute the FDR and control the genes than edgeR. I ran analysis with the same initial filtering applied in edgeR and further used the same filtered data for DESeq2 with cooksCutoff=F, independentFiltering=FALSE in results function.

Parameters Used in both methods are:

rowSums(cpm(data)>1) >= ncol(data)/2

Sample = 6; each have 3 replicates ; group = C and T

adjust.method = "BH"

Dispersion , Model fit and testing in both methods applied = GLM and LRT

In particular for edgR:

y <- estimateGLMCommonDisp(y,design)

y <- estimateGLMTrendedDisp(y,design)

y <- estimateGLMTagwiseDisp(y,design)

fit <- glmFit(y,design)

lrt <- glmLRT(fit,coef=2)

In particular for DEseq2:

ddsLRT <- DESeq(dds, test ="LRT", reduced = ~1)

resLRT <- results(ddsLRT, cooksCutoff=F, independentFiltering = F)

Results:

No. of DE genes detected in edgeR ( Raw p-value <= 0.05) = 16 # I am only showing here 9 genes/miRNAs.

No. of DE genes detected in edgeR ( FDR <= 0.05) = 8

No. of DE genes detected in edgeR ( FDR <= 0.05) = 4

Note: Basically i am wondering about last 5 genes/miRNAs. Any clue ??

Results from edgeR and DESeq2

RNA-Seq R DESeq edgeR miRNASeq • 2.2k views
ADD COMMENT
1
Entering edit mode
8.6 years ago

They use the exact same function to compute the adjusted p-value. Since this is dependent on the distribution of p-values and the p-values themselves are slightly different, this sort of result is typical.

ADD COMMENT

Login before adding your answer.

Traffic: 2593 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6