[Bioc-sig-seq] Shortreads

Martin Morgan mtmorgan at fhcrc.org
Wed Aug 27 18:14:53 CEST 2008


Hi Terry --

Terry Therneau <therneau at mayo.edu> writes:

>  
>   Thanks for all the responses.
>
>   I have successfully put R-devel up on a machine, and then installed the 
> shortread package.  To start, I was trying to have a local library with the 
> add-ons, working with the installed R base on a shared server.  This was not a 
> straightforward exercise.
>   
>   A suggestion: much of my confusion was due to the fact that there were not any 
> manual pages for biocLite, bioinstall, etc; or at least no documentation that I 
> easily found.  So I tried the standard R install (no luck), and various other 

The horse is out of the barn, but there is some documentation at

http://bioconductor.org/install

The key to making 'install.packages' work is to use the appropriate
repositories, available with biocinstallRepos() once biocLite.R has
been sourced. The idea of a message, e.g., 'See
http://bioconductor.org/install for help' sounds good.

> fruitless paths based on wrong guesses.  Sean's explanation of why biocLite does 
> what it does was helpful; I'd already guessed some of it by reading source code. 
> Perhaps the standard install of biocLite could finish with a message pointing 
> one to  a short 1-2 paragraph overview.  When it works biocLite is nice, but 
> when it doesn't there isn't much hint of where to begin.  It consistently 
> ignored my "lib" directive, for instance.  And from my reading of
> the source

Hmm, more detail (off list, perhaps) might help, but the 'lib'
argument gets passed to install.packages as expected, for me
(sessionInfo() snippet below).  install.packages behavior is not
trivial when lib is provided.

(My own approach is to set an environment variable R_LIBS. This gets
incorporated into .libPaths(), and packages are installed and read
from a consistent location across sessions, without explicitly
providing a 'lib' argument to biocLite).

> code it wouldn't have helped in the end if it had paid attention -- there is, is 
> there not, an assumption that the "stubs" are in an include directory in the R 
> home tree, not the user's?
   
I don't think so. For instance, if IRanges is installed in
R_HOME/library and Biostrings in /tmp/mydir, then ShortRead is
compiled with appropriate paths -- along the lines of

gcc -std=gnu99 -I/home/mtmorgan/arch/x86_64/R-devel/include  \
    -I/usr/local/include \
    -I"/home/mtmorgan/arch/x86_64/R-devel/include/IRanges/include" \
    -I"/tmp/mydir/Biostrings/include" \
    -fpic  -g -O2 -c alphabet.c -o alphabet.o

It's complicated, though.

> sessionInfo()
R version 2.8.0 Under development (unstable) (2008-08-22 r46416) 
x86_64-unknown-linux-gnu 

>   I anticipate that the Solexa, and the threat of a couple more coming in soon, 
> will accelerate my familiarity with bioconductor.  The shortread package already 
> has repaid with interest for these install nuisances.

Always good to hear!

Martin
   
>   	Terry T.
>
> _______________________________________________
> Bioc-sig-sequencing mailing list
> Bioc-sig-sequencing at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioc-sig-sequencing

-- 
Martin Morgan
Computational Biology / Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N.
PO Box 19024 Seattle, WA 98109

Location: Arnold Building M2 B169
Phone: (206) 667-2793



More information about the Bioc-sig-sequencing mailing list