Significant genes in DESeq2 analysis
Entering edit mode
cbio • 0
Last seen 24 days ago

Hi everyone,

I have a theoretical question related to results after p-value adjusting methodology.

After a classical RNAseq experiment with one condition, processing counts (filtering small counts in at least half of the samples), normalizing the information, etc. I'm QCs the information, plotting the distribution of p-values of all the genes and then splitting then the information by the significance or not using a p-adj < 0.1. After reading some information on how to interpret the information here or here, I was expecting a threshold in the p-value that would more or less equal the Null hypothesis to the noise of the different p-values.

My questions are:

  • Is it normal to obtain significant genes in all that have a p-value < 0.01 (first bar)? Shouldn't some noise be identified?
  • Shouldn't some differentially expressed genes be identified in the first 4 or 5 bars that clearly should have some positives?
  • Am I losing possible true positive results and should try with other methodologies/parameters?

Thanks in advance to all.

First experiment

I have a different case, with more samples, and although the clusterisation in PCA was worse, there's still what seems to be a good pattern in the p-value distribution. But only the lower p-value are detected as significant:

Different experiment

DESeq2 RNASeq • 181 views
Entering edit mode
Last seen 7 days ago
United States

The short answer is that your results are fine. There's no expectation for what proportion of bars in a histogram of the p-values should be significant using an FDR>0.1.

Login before adding your answer.

Traffic: 736 users visited in the last hour
Help About
Access RSS

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

Powered by the version 2.3.6