Skip to main content
Topic: P-Value of zero in common features? (Read 4612 times) previous topic - next topic

P-Value of zero in common features?

I am observing a p-value of zero in my metaXCMS common features? It is obviously incorrect to report this p-value..Is there a certain number of decimal places which metaXCMS reports? Are these values indeed significant or not? Has anyone else seen this?

Thanks in advance.  :)  :)

Re: P-Value of zero in common features?

Reply #1
metaXCMS does not modify the p-values of your experiments.
Do you have p-values of zeroe in the diffreports that you import into metaXCMS ?

Re: P-Value of zero in common features?

Reply #2
Hi Ralf,

I have checked through the diff reports and the highest P-value for feature lowest ranked feature in each subject is 0.98 or so. There are no 0 values? The highest ranked common feature is 166.021416 m/z with the following results: Subject A (control vs. intervention) UP(fc=3.7,pval=0), Subject B (control vs. intervention) UP(fc=2.3,pval=0.00018), Subject C (control vs. intervention) UP(fc=3.1,pval=0). I have checked for this feature in each individual diff report and I get the following p-values Subject A  (p=5.77E-07), Subject B (p=0.000184097), Subject C (p=2.51E-06). Any idea why I might be getting zero values like this in the common features? I am filtering by 0.01 p-value, 2 fold change, then m/z tolerance of 0.005 and an RT tolerance of 20 seconds? Don't know if this clarifies it is not in all of the common features just some.

Thanks for your help,

Will  :)

Re: P-Value of zero in common features?

Reply #3
The p-value is rounded to 5 digits using R's round() function.
It seems that round() just outputs zero for very small numbers.

The p-value of that particular feature must be very small, have you tried looking it up in the original diffreport for subject C ?

I guess I could make the number of digits an option.