Question: alpha and p-adjust value
0
gravatar for quark.vera
8 months ago by
quark.vera0
quark.vera0 wrote:

Hello;

I am using DESeq2 for DEG discovery, however I found changes in p-adjust value depending the alpha I used in the same gene. For example using the default alpha value (0.1) for results I got:

res05 <- results(ddi, contrast=c("treatmenttissue",conditionsName[2],conditionsName[6])) res05Ordered <- res05[order(res05$pvalue),] data05 <- na.omit(as.data.frame(res05Ordered)) keepAllDEgenes <- (data05$padj<=0.05) genesDE<-data05[keepAllDEgenes,] genesDE["TRINITYDN64996c0g2",] baseMean log2FoldChange lfcSE stat pvalue TRINITYDN64996c0g2 31.79655 -1.443795 0.4637561 -3.113262 0.001850314 padj TRINITYDN64996c0g2 0.04894679

But if I change to alpha=0.05

res05 <- results(ddi, contrast=c("treatmenttissue",conditionsName[2],conditionsName[6]),alpha=0.05) res05Ordered <- res05[order(res05$pvalue),] data05 <- na.omit(as.data.frame(res05Ordered)) data05["TRINITYDN64996c0g2",] baseMean log2FoldChange lfcSE stat pvalue TRINITYDN64996c0g2 31.79655 -1.443795 0.4637561 -3.113262 0.001850314 padj TRINITYDN64996c0g2 0.2324749

No my gene is not significan differential expressed in means of p-adjust values. This become annoying beacause if now I did a cut-off alpha of 0.01

res05 <- results(ddi, contrast=c("treatmenttissue",conditionsName[2],conditionsName[6]),alpha=0.01) res05Ordered <- res05[order(res05$pvalue),] data05 <- na.omit(as.data.frame(res05Ordered)) data05["TRINITYDN64996c0g2",] baseMean log2FoldChange lfcSE stat pvalue TRINITYDN64996c0g2 31.79655 -1.443795 0.4637561 -3.113262 0.001850314 padj TRINITYDN64996c0g2 0.0378142

This same gene become a DEG. My question is why do the padjust value changes drastically depending an alpha value?, and if so why 0.05 is worst than 0.01 in means of p-value tests.

I hope you can answer.

Best

Arturo.

deseq2 • 233 views
ADD COMMENTlink modified 8 months ago by Michael Love25k • written 8 months ago by quark.vera0
Answer: alpha and p-adjust value
1
gravatar for Michael Love
8 months ago by
Michael Love25k
United States
Michael Love25k wrote:

Yes, if you read the help page for ?results, you will see that alpha changes the target for the independent filtering procedure, therefore producing different adjusted p-values. See also the Independent Filtering section of the vignette.

ADD COMMENTlink written 8 months ago by Michael Love25k

Hi Michael, Thank you. I'll take a look to filtering procedure section.

ADD REPLYlink written 8 months ago by quark.vera0
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 255 users visited in the last hour