rietho / IPO

A Tool for automated Optimization of XCMS Parameters
http://bioconductor.org/packages/IPO/
Other
34 stars 20 forks source link

MacOSX: Different results for xcmsSet.centWave.mzdiff #42

Closed lecorguille closed 7 years ago

lecorguille commented 8 years ago

Hi, Because I'm working on the IPO Galaxy wrappers. I set some functional tests so I often launch IPO with the same set of inputs and parameters.

First of all, note that I get this behaviour on a Mac.

And maybe it's completely normal but I get different results for this value xcmsSet.centWave.mzdiff sometimes I get 0.01 and sometimes I get -0.001.

I thought that was because I was loading or not snow at the beginning of my script? But I also got this switch when I just add or not a print(sessionInfo()) before IPO functions.

Meanwhile, it is stable at 0.01 under Linux and through my TravisCI. So it's not a big deal because Linux is my target. But I would like to share this experience.

rietho commented 8 years ago

Thanks for sharing your experience.

That's an interesting behavior. I'd expect that the PPS value for your case is the same for both xcmsSet.centWave.mzdiff outcomes. Is that true? In this case I'd believe that there are some functions underlying some randomness on Mac, but not on Linux.

lecorguille commented 8 years ago

Sorry, but what is PPS?

rietho commented 8 years ago

The PPS is the Peack-Picking-Score that is used as objective for the peack-picking optimization (see our paper for more details).

IPO is now also available through the development version of Bioconductor. In the vignette on Biocondcutor you can see how to get the PPS value of an optimized peak-picking (just search for PPS xset in the vignette).

lecorguille commented 8 years ago

I can't find this new version of the vignette into this repo. Is there an other place for IPO code?

rietho commented 8 years ago

This repo will stay the the place for the IPO code. I have not pushed the preparation for bioconductor before, but now there is a branch called biocprep with all preparations for biocondcutor. Nevertheless the vignette here is not knitted. But you can find the vignette with all R output on bioconductor development version http://bioconductor.org/packages/devel/bioc/html/IPO.html