PMDDA Error in unserialize(node$con) : error reading from connection

[wushaowen1992]

Hi yufree,
I am trying to use this good tool xcmsrocker that you developed and try to repeat your PMDDA workflow, however, when I run the demo data, and reach to the step:

IPO

resultRetcorGroup <-
optimizeRetGroup(xset = optimizedXcmsSetObject,
params = retcorGroupParameters,
plot = F,
subdir = NULL)

I always got an error says "Error in unserialize(node$con) : error reading from connection".
From the internet, it seems to be a problem with parallel calculation, I wonder if you know how to deal with this issue?
Thanks very much for your help.

[yufree]

Yes. This error is parallel computation related. I wonder where you run the xcmsrokcer? HPC or local computer?

[wushaowen1992]

I am runing on a local computer, with 16 cores of CPU and 32 G of memory.

[wushaowen1992]

How can I reduce the number of parallel calculation in order to keep it going? Although this will prolong the running time.
Thanks.

[yufree]

You can add the following code before the error codes to skip the parallel computation:

BiocParallel::register(BiocParallel::SerialParam(), default = TRUE)

Another option is to use a local version of PMDDA. xcmsrocker is mainly designed for hpc and user need to know how to mapping local path through docker to use their own data. In your case, a local version might be much easier. You could clone this repo: https://github.com/yufree/pmdda

Click 'PMDDA.Rproj' in RStudio to open the project. Then you will need renv package to reproduce the software environment needed for PMDDA:

install.packages('renv')
renv::restore()

[wushaowen1992]

Thanks for this detailed information. I will try to follow the local version.

[wushaowen1992]

Hi,
I tried to work with both xcmsrocker and local PMDDA project, both have some issues.
I tried to use xcmsrocker, with BiocParallel::register(BiocParallel::SerialParam(), default = FALSE), IPO could work.
but when I reach to the step, # back up peak list as csv file and xcmsSet object
mzrt <- enviGCMS::getmzrt(srm, name = 'srm')
I see the srmxset.rds, but no csv file, and an error says:
Note: you might want to set/adjust the 'sampclass' of the returned xcmSet object before proceeding with the analysis.
Error in (function (classes, fdef, mtable) :
unable to find an inherited method for function ‘featureValues’ for signature ‘"xcmsSet"’
How could I deal with this issue?
Thank you.

[yufree]

It's wired. Anyway the csv file could be generated via the srmxset.rds. You can try the following code:

mzrt<-readRDS('srmxset.rds')
enviGCMS::getmzrt(mzrt,'srm')

Then you should see srmmzrt.csv.