[BioC] Does old workspace work?
engrav at u.washington.edu
Sat Nov 15 06:18:21 CET 2008
So single threaded
Not worry about warnings
64bit R.app installs 64bit stuff
Rm'd the old scripts
> From: Kasper Daniel Hansen <khansen at stat.berkeley.edu>
> Date: Mon, 10 Nov 2008 20:30:57 -0800
> To: Loren Engrav <engrav at u.washington.edu>
> Cc: "bioconductor at stat.math.ethz.ch" <bioconductor at stat.math.ethz.ch>
> Subject: Re: [BioC] Does old workspace work?
> On Nov 10, 2008, at 11:50 , Loren Engrav wrote:
>> So I fixed up the old workspace from early 2007 with all your help
>> Installed 64bit Intel MacOS stuff
>> Installed the new BioC maanova stuff
>> Changed one of the 60 chips
>> And reran the mixed linear regression previously done in early 2007
>> The command was
>> ftest.Breed.mix_1000perm081109 <- matest
>> term = ³Breed², n.perm = 1000, critical = .9, test.type = c("ftest"),
>> shuffle.method = c("sample"), MME.method = c("REML"), test.method =
>> c(1,0,1,1), pval.pool = TRUE, verbose = TRUE)
>> In early 2007 this process took about 40 hours
>> This time it took 15 hours which is a serious improvement
>> Four derivative questions Please
>> #1I noticed the computer was using only one of the 8 processors
>> Can R/BioC use all of the processors?
> In general R is single threaded, so no. However, on Mac OS, R is
> (typically) build against "Altivec" which is an Apple supplied,
> multithreaded version of BLAS and Lapack. What this means is that any
> computation that involves matrix algebra, will use all your cores.
> This could very well be behind the substantial speed up you see. It is
> also possible to make packages utilize multiple cores, but this is
> somewhat non-standard and hard to do. One example is affypara, for
> preprocessing of affy arrays.
>> I received this warning many many times
>> 50: In any(parsed.formula$random) ... : coercing argument of type
>> to logical
>> #2Is this significant?
> I have no idea.
>> I have 64bit R.app but to install the 64bit packages I quit R.app
>> and in
>> terminal do
>> source ("http://www.bioconductor.org/biocLite.R")
>> biocLite (c("whatever"), type="source")
>> It seems I cannot install the 64bit packages from R.app
>> #3Is this true? This has been discussed but I cannot find the
> If you are using Mac OS X on an Intel machine, the latest Bioconductor
> builds includes 64bit versions. So that means that type = "mac.binary"
> should give you a 64 bit version.
> Whether or not you do this from R.app should not have any impact.
>> In the old workspace there are several objects like
>> which are scripts of various types I guess to install groups of
>> contriburl = "http://bioconductor.org/packages/1.9/bioc/src/contrib
>> available.packages(contriburl)[, "Package"]
>> #4Are these scripts old and no longer used to install groups?
> These scripts are by-product of doing
> R> source(http://www.bioconductor.org/biocLite.R")
> and can safely be deleted. They will be recreated, if necessary,
> everytime you source biocLite.
>> Thank you again
>>> From: Loren Engrav <engrav at u.washington.edu>
>>> Date: Fri, 31 Oct 2008 15:38:12 -0700
>>> To: "bioconductor at stat.math.ethz.ch" <bioconductor at stat.math.ethz.ch>
>>> Subject: Re: [BioC] Does old workspace work?
>>> Cool, I can deal with evolution
>>> And the archives will be a super addition to ? and google, thank
>>> you, maybe
>>> better than google
>>> So we biopsied shallow and deep wounds on 3 Duroc pigs and 3
>>> Yorkshire pigs
>>> at 1 2 3 12 and 20 weeks, this then is 60 porcine chips with 24123
>>> sets on the chip, we were interested in chip2-chip1, chip4-chip3, etc
>>> And studied the differences with R/maanova and various biologic data
>>> reduction steps to achieve a group of 1289 for further study
>>> So for the fun of it, I compared chip2-chip1 in gcrmaOld and
>>> In gcrmaOld there were no differences of exactly zero
>>> In gcrmaNew 249/24123 were exactly zero which seems kinda funny
>>> This affected 4 in the group of 1289 but did not change anything
>>> 9247/24123 differences changed sign which is potentially bad for
>>> this study;
>>> but after cutting those with log ratio < .5 as too trivial to worry
>>> about in
>>> this complex system, zero were left
>>> I could rerun this, but given this, and that the old Mac PRO PPC
>>> took ~40
>>> hours to run the mixed linear regression, I think the old data is
>>> just fine
>>> And since
>>> objNew <- as(objOld, "ExpressionSet") works would appear I am back in
>>> Thank you all
>>> From: Sean Davis <sdavis2 at mail.nih.gov>
>>> Date: Fri, 31 Oct 2008 15:00:47 -0400
>>> To: Loren Engrav <engrav at u.washington.edu>
>>> Cc: "bioconductor at stat.math.ethz.ch" <bioconductor at stat.math.ethz.ch>
>>> Subject: Re: [BioC] Does old workspace work?
>>> On Fri, Oct 31, 2008 at 2:47 PM, Loren Engrav <engrav at u.washington.edu
>>>> Thank you, got it
>>>> exprs(obj) is missing from the output but shows in ?ExpressionSet
>>>> and puts
>>>> up the data so I can see it
>>>> And now perhaps one final question to understanding this old data
>>>> Long long ago I did
>>>> AA_ReadAffy <- ReadAffy()
>>>> ABeset_gcrma < gcrma (AA_ReadAffy)
>>>> Now I do
>>>> AA_ReadAffy081031 <- ReadAffy() #on the same .cel files
>>>> ABeset_gcrma081031 <- gcrma (AA_ReadAffy081031)
>>>> Then I do
>>>> Show(ABeset_gcrma) and exprs(ABeset_gcrma081031)
>>>> And they do not match, for example
>>>> Probe set Ssc.10026.1.A1_at
>>>> ABeset_gcrma is 4.124689
>>>> ABeset_gcrma081031 is 3.344443
>>>> ReadAffy changed?
>>>> gcrma changed?
>>> The bioconductor archive is your friend.
>>> Bioconductor mailing list
>>> Bioconductor at stat.math.ethz.ch
>>> Search the archives:
>> Bioconductor mailing list
>> Bioconductor at stat.math.ethz.ch
>> Search the archives:
More information about the Bioconductor