Adjusted p-values approaching minimum value of double

[Cailin Jordan]

Hi,

I am running an analysis using a small number of ATAC-seq samples (2 distinct replicates for one condition, one merged file for the other). I don’t have much control over this since I am using previously generated data. When I run diffTF in analytical mode, I get what looks like a pretty reasonable volcano plot, except for the fact that the top hits have the same, very small adjusted p-value (<1e-300). It seems from the plot and summary that the adjusted p-value might be the minimum value of a double. The p-values are also all identical for these TFs. Is there a way for me to adjust the parameters so I can get valid p-values for all of the TFs?

[Profound Profundity]

Hi,

Just another user here. I have a few questions about how you are runnign the program since I am having errors that I have posted here.

• I was able to run the program on python, but now I am getting errors.

1. Are you using Linux or Windows?
2. Singularity version gives me a bash shell error – are you using singularity?
3. Now even the python version is erroring out.

,,,,,, can you please share how wre running the program?

‌

Thanks in advance!

‌

[Cailin Jordan]

Hi, I’m running on linux with Singularity installed - I installed both Singularity and DiffTF according to instructions in the documentation. I was never able to get the locally installed version of the program to run successfully.

‌

[Christian Arnold]

Hi there, if using the analytical mode, p-values can indeed become very small to the point that they are smaller than the minimal non-zero value that can be represented. That is why diffTF sets such p-values to the the minimum non-zero value. In your case, the differences, despite just having few samples, seem to be so large that this happens for multiple TFs. In the end, not the adjusted p-value is critically important as long as it is small and significant according to how significance is quantified, but the TF activity (x-axis) difference is more important.

[Cailin Jordan]

Thank you for clarifying!