WGCNA package FAQ

Installation problems

1. R 2.10 complains about package being built under R-2.7.
   
   The authors of R apparently changed the way help files are compiled. We will do our best catch up, but in the meantime please ignore the error. The package is usable and even help seems to work alright as far as we can tell.
   
2. R complains about wrong package type.
   
   A common cause of this error is that when the user saves the file, the operating system will uncompress or unzip it. Typically, this means the .zip or .tar.gz bundle will be decompressed and extracted, which renders the file unusable for R. For example, Mac OS X seems to automatically decompress the gzipped file. The solution is to save the file to disk as is, without letting any program such as WinZip touch it. R will decompress and unpack the package itself. On a Mac, you may have to open a terminal, change to the directory where you saved the file, and type
   
   gzip WGCNA_*.tar
   
   
3. The package won't install on my Mac.
   
   Please look at the installation instructions and make sure you have the required XCode tools installed.
4. I have Xcode tools installed and the package still won't install.
   
   Some users have reported that a package names gfortran.pkg is also necessary. This may be a new feature of R as of version 2.9.0.

Runtime errors

1. I keep getting a malloc error.
   
   Several Mac users have reported malloc errors such as
   
   R(9073,0xa013dfa0) malloc: *** mmap(size=95006720) failed (error code=12)
*** error: can't allocate region
*** set a breakpoint in malloc_error_break to debug

   
   From all we know this is a spurious and harmless message, and can be ignored. For all C- and gcc-savvy Mac users out there, if you find a cause and solution, please let us know.
2. The code crashes with a mysterious error message.
   
   Such crashes are often the result of unusual data that cause a condition we have not thought about, or subtle abnormalities in the data. Here are a few situations in which we have seen our code crash:
   • The data contains too many missing entries. We attempt to make the code resistant to missing data, but sometimes we miss something. As a test, try imputing the missing data and run a test run with imputed data.
   • The data may contain too few samples or probes.
   • The input data is not numeric. This often a subtle and hard to catch problem. When reading data in from text tables (tab-separated .txt or comma-separated .csv files), R may convert a data frame to a list and/or convert some columns to a character or a factor. This may happen, for example, because missing data in your data file are encoded as NULL or N/A, while R expects NA and everything else is treated as a character string. It is best to sort such problems out at the source, but it may not always be possible, and the user will need to run a version of
     

       charData = apply(as.matrix(data), 2, as.character); 
       numData = apply(charData, 2, as.numeric);
     

   • Last but not least, your data may be perfectly fine but lead to no modules, or perhaps just one module, and such cases may not be handled correctly by the code (i.e., the code is buggy). Please send Peter Langfelder an email if you encounter such a situation.