AffyExtensions - Older versions


Older versions of the package AffyExtensions which contains a set of routines that can be called from within R.
Version 0.3 series requires affy (Version 1.0.2) and bioconductor 1.0 release
Linux/Unix and the likeWindows
Version 0.4 series requires affy (Bioconductor 1.1 release)
Linux/Unix and the likeWindowsDateComments
Versions 0.4-1 through 0.4-14 are available upon request here
AffyExtensions_0.4-15.tar.gzAffyExtensions_0.4-15.zipFeb 27, 2003Incremental updates. Slightly longer vignette. Except for bug fixes, the last release in the 0.4 series
Version 0.5 series requires affy (Bioconductor 1.2 release)
Linux/Unix and the likeWindowsDateComments
Versions 0.5-0 through 0.5-13 are available upon request here
AffyExtensions_0.5-14.tar.gzAffyExtensions_0.5-14.zipSep 12, 2003Fix a bug in how constraints are applied
Version 0.6 series requires affy (Bioconductor 1.3) package. Note that 0.6 is the development series
AffyExtensions_0.6-4.tar.gzNot currently availableJan 12, 2004PLM.tstat introduced
AffyExtensions_0.6-5.tar.gzAffyExtensions_0.6-5.zipMar 13, 2004Interaction style terms can now be specified in model for fitPLM
AffyExtensions_0.6-6.tar.gzNot currently availableApr 29, 2004PLM.tstat/PLM.fstat introduced
AffyExtensions_0.6-7.tar.gzNot currently availableMay 24, 2004Fixed some memory leaks. Compatible with BioC 1.4 and R 1.9.X

Longer history of changes for current (0.6) series

VersionDate of ReleaseComments
0.6-0Never released publically (only available to UCB)

the PLMset object now has new attributes

  • residuals - a matrix storing residuals from model fit (notice that these are
  • normalization vector -vector (in sorted order of PM probe intensities for normalization distribution)
  • DESCRIPTION item is now a list, that can be used more carefully to describe how the model was fit
  • residual standard error and df
  • varcov - variance covariance matrices for each model

"residuals", residual SE, varcov are returned in the PLMset object returned from fitPLM (depending on user flags).

image() on a PLMset can now plot residuals and possible to add legend bars.

Some major changes in how parameters are passed in some of the C code routines in the fitPLM process (this should not be of consequence to the user)

boxplot function for exprSets

normalize.exprSet.invariantset - now has same choices of baselines as in the affy package.

conversion function pset2eset() which converts a PLMset to an exprSet

fitPLM now has parameters output.param and model.param which are lists. It allows the user better control over how model is fit and what is outputted.

rmaPLM() function lets you fit the rma model as a PLMset. This allows you to use the image function on residuals.

threestep() now has a summary.param parameter which gives you some control over the summarization routines.

threestepPLM() function returns residuals in addition to expression values in a PLMset

0.6-1Oct 18, 2003Preliminary support for binary CEL files. Note that this package works with Release 1.2 bioconductor and R-1.7.1
0.6-2Nov 20, 2003Minor fixes. Support for R-1.8.0 and Release 1.3 Bioconductor
0.6-3Dec 8, 2003Remove all cel file parsing code. This is now in the base affy package (1.4.3 or later). Replacement functions for residuals/resid.
0.6-4Jan 12, 2004sampleNames replacement function for PLMset objects
weights/residuals allow you to index probes from specific probesets
resid/residuals functions allow you to choose standardized residuals
a general test statistic function PLM.tstat
0.6-5Mar 13, 2004 Change how kappa is computed in SE algorithm (this affects non-Huber psi fns)
threestep,rmaPLM,threestepPLM,fitPLM - accept the subset parameter which allowing you to restrict computations to a set of specified probesets.
MAplot() function gives n MA plots equivalent to Mbox()
Interaction style terms can now be specified in model for fitPLM
0.6-6Apr 29, 2004PLM.tstat has been updated.
PLM.fstat introduced
Some of the man/ documentation has been cleaned up
A function PLM.flags() has been introduced
0.6-7May 24, 2004Fixed some memory leaks. Compatible with BioC 1.4 and R 1.9.X

Longer history of changes for 0.5 series

VersionDate of ReleaseComments
0.5-0Mar 27, 2003Initial Release. New background methods (not default), ability to do MAS like background. Only version of the 0.5 series that will work with versions of R earlier than R-1.7.0
0.5-1May 13, 2003Requires R-1.7.x and affy version 1.2.19 or later. This means the package should work well with the forthcoming Bioconductor 1.2 release. The default background method in fitPLM and threestep has been changed so that it agrees with affy. Note that some lapack routines are now used internally in preference to linpack.
0.5-2May 30, 2003Another normalization method called quantiles.probeset. Currently it is a separate function, it will be integrated into threestep() and fitPLM() at a later point.
0.5-3Jun 11, 2003Other M estimation Psi functions have been added to fitPLM(). These may be specified using the psi.type parameter. (An additional parameter psi.k controls a tuning parameter). Currently iteration of all of these methods is started at the ordinary linear regression estimates.
0.5-4Jun 12, 2003A function callPA was introduced. It makes Presence/Absence calls. Note that this is not the same method used by the MAS 5.0 software
0.5-5Jun 15, 2003First version of read.affybatch2. It is designed to be faster with large numbers of chips. Overload your read.affybatch by using read.affybatch <- read.affybatch2 to use it with ReadAffy()
0.5-6Jun 17, 2003Modifications to read.affybatch2. It now deals with OUTLIER and MASKS sections if required. Windows build uses R-1.7.1
0.5-7Jun 19, 2003Modifications to read.affybatch2. This should cut a little more time off the run time of this routine.
0.5-8Jun 24, 2003You can now call "quantile.probeset" as a normalization method in threestep(), fitPLM()
0.5-9Jul 19, 2003Mbox function. Note that the parsing routines are now also available in the affy devel packages.
0.5-10Jul 19, 2003fix error in how constraints are applied.
0.5-11Aug 20, 2003
SE are now outputed from some threestep methods.
Reduce memory overhead in threestep.
Fix a bunch of compiler warnings (all were not fatal) that have not been worked on previously.
A vignette describing the threestep function.
Changes to how options to background and normalization methods are handled in threestep/fitPLM.
normalize.AffyBatch.quantiles.probeset can deal with MM values.
callPA() has been modified, it provides more options.
0.5-12Aug 26, 2003
Normalization routines for exprSets. There is now a generic function normalize() that may be applied to exprSets in the same manner as AffyBatches.
A function for doing the scaling normalization added (differs slightly from normalize.constant in that pseudo baseline chips may be used and a trimmed mean might be used). can call via normalize.AffyBatch.scaling, normalize()
changed normalize.quantiles.probeset so it can now work on natural scale
0.5-13Aug 28, 2003
fitPLM/threestep can now send proper parameters to quantiles.probeset.
a little more control over what happens to an affybatch when you use the scaling normalization.
0.5-14Sep 12, 2003
There was a bug in how constraints were applied. If the variable was defined in the parent environment then the constraint was not properly applied.

Longer history of changes for 0.4 series

VersionDate of ReleaseComments
0.4-0Jan 9, 2003Initial Release. Not all background or summary methods included.
0.4-1 Jan 11, 2003More Summary methods, background methods working. Calling methodology refined
0.4-2 Jan 13, 2003Introduction of robust linear model in threestep method.
0.4-3 Jan 22, 2003Introduction of PLMset object. fitPLM() function introduced, only fits default model
0.4-4 Jan 23, 2003Ability to make image plots of the weights
0.4-5 Jan 27, 2003Wider set of models may be fitted. First pass attempt at standard errors.
0.4-6 Jan 28, 2003Better attempts at standard errors. Please add se.type=4 to your fitPLM calls, default se.type is numerically unstable
0.4-7 Jan 30, 2003Fix to se.type=4 (which has become the default type). Distinct row names on weights matrix (to allow write.table to work better). A boxplot type function for PLMset objects
0.4-8 Feb 2, 2003Fix so that se.type=2 and se.type=3 should work properly now (numerical instability removed). Some slight optimizations.
0.4-9 Feb 6, 2003Incremental updates to fitPLM. A vignette. Four new summary methods in threestep.
0.4-10 Feb 9, 2003Optimization to se.type=4 in fitPLM. Fuller vignette.
0.4-11 Feb 11, 2003A number of space, time optimizations. Fix to se.type=1,2,3 (can explain why differ from summary.rlm)
0.4-12 Feb 15, 2003Incremental updates. Vignette should work with windows build.
0.4-13 Feb 18, 2003Ability to fit models with intercept coefficients and to fit models with out probe effects. Accessor functions for probe/ intercept parameters
0.4-14 Feb 25, 2003Ability to fit models with sum to zero constraints. image gives correctly oriented plot.
0.4-15 Feb 27, 2003Incremental updates. Slightly longer vignette. Except for bug fixes, the last release in the 0.4 series

Older Documentation

Documentation for AffyExtensions (0.3-3) is in this PDF file.

Special Notes

This software as of version (0.3-3) is recommended only for use with affy 1.0.2. Release 1.1 of Bioconductor brings with it a much changed affy package. A fast RMA has been integrated into that release.