BSseq() will no longer reorder inputs. Previously, the
returned BSseq object was ordered by ordering the loci, although
this behaviour was not documented. BSseq() may still filter out loci
if rmZeroCov = FALSE or collapse loci if
strandCollapse = FALSE or duplicate loci are detected, but the
relative order of loci in the output will match that of the input.
Fix bug with maxGap argument of BSmooth(). The bug
meant that the 'maximum gap between two methylation loci' was incorrectly
set to 2 * maxGap + 1 instead of maxGap. This likely did not
affect results for users who left the default value of maxGap = 10^8
but may have affected results for small values of maxGap.
Cleaning up Imports and Suggests.
1.13.6: Fix performance regression in BSmooth(). Thanks to
Shan Andrews for the report
(https://github.com/kasperdanielhansen/bsseq/pull/57).
1.13.5: Fix major bug in combine() and combineList().
This bug led to bad BSseq objects with incorrect methylation
estimates due to incorrect 'M', 'Cov', 'coef', and 'se.coef' assays. To be
safe, BSseq objects created with versions 1.13.0 to 1.13.4 should be
re-created using a newer version. More specifically, any BSseq objects
created with combine() or combineList() should be re-created.
Also, BSseq objects created using read.bismark() or
read.bsmooth() with multiple 'files' should be re-created.
Thanks to Alejandro Reyes (@areyesq89) for the report
(https://github.com/kasperdanielhansen/bsseq/pull/54).
1.13.4: Fix performance regression in getMeth() and
getCoverage() when 'regions' were supplied. Thanks to Alejandro
Reyes (@areyesq89) for the report
(https://support.bioconductor.org/p/97611/).
Moved vignettes to Rmd.
bsseq now uses DelayedMatrix objects from the DelayedArray package for all matrix-like data. This enables large data to be stored on disk rather than in memory.
Serialized (saved) BSseq, BSseqTstat, and BSseqStat objects will need
to be updated by invoking x <- updateObject(x).
Fixing an error with reading multiple samples aligned with bismark in the format "cytosineReport".
new function strandCollapse for collapsing forward and reverse strand data to be unstranded.
Updated read.bismark() to support the cytosine report files; both formats are supported. Other minor updates (mostly internal) to read.bismark(). Greatly improved documentation of this function, paying particular attention to differences in file formats between versions of Bismark.
Converted to using Authors@R in the DESCRIPTION file.
plotRegion did not respect the col/lty/lwd argument if given explicitely as opposed to through pData(). Reported by Karen Conneely <[email protected]>.
Fixed an issue introduced by the previous change (to plotRegion). Reported (with a fix) by Tim Triche Jr <[email protected]>.
Fixed a serious bug in collapseBSseq reported by Julien Roux <[email protected]>: it would use the Meth values instead of Cov for the final object's Cov. However, this will result in the return object having a methylation of 100 percent across all loci, so hopefully users will have seen this for themselves.
Fixed a bug in combineList which made combineList use "slow" code even in a special case where a faster shortcut was possible. This bug fix does not change the output of the function under any input, it only makes it faster. Reported by Julien Roux <[email protected]>.
validity now checks for the presence of colnames (sampleNames) which was assumed to be set. Reported by Kevin Rue-Albrecht <[email protected]>.
Fixed a man page issue.
Slightly changed CITATION.
Added GitHub URL to DESCRIPTION.
Fixed a problem with "width" in the title of bsseq plots.
plot.BSseqTstat now allows for BSseqTstat objects computed without correction.
validObject(BSseq) has been extended to also check for sampleNames consistency.
Fixed a bug related to validity checking.
Increased maxk from 10,000 to 50,000 in calls to locfit, to allowing fitting the model on genomes with unusally long chromosomes (Thanks to Brian Herb for reporting).
The class representation for class 'BSseq' has changed completely. The new class is build on 'SummarizedExperiment' from GenomicRanges instead of 'hasGRanges'. Use 'x <- updateObject(x)' to update serialized (saved) objects.
Fixing a problem in orderBSseq related to chromosome names.
Allowed user specification of maxk, with a default of 10,000 in BSmooth.
Many bugfixes made necessary by the new class representation.
Better argument checking in BSmooth.tstat.
A few undocumented functions are now documented.
Rewrote orderBSseq
Removed the returnRaw argument to read.bismark() as it was unnecessary (Bismark output files does not have additional information beyond M and Cov and genomic positions, unlike BSmooth).
Moved the Bismark example data from data to inst/extdata.
combineList() now deals with the case where the list of BSseq objects have different genomic locations. This speeds up read.bismark() substantially.
Exposed combineList() as a faster alternative to Reduce(combine, list).
Updated the code for the plotting routines (plotRegion). This should not have an impact on user-visible code.
Added read.bismark() function to parse output from the Bismark alignment suit [thanks to Pete Hickey].
Refactorized plotting code.
Fixed a bug in getMeth, where type="raw" resulted in an error for non-smoothed data objects.
Updated CITATION and citations in the vignettes.
Now read.bsmooth supports both gzipped and non-gzipped files, whereas previously it assumed the output files to be gzipped. Thanks to Andreas Schoenegger for reporting this problem.
Fixed a bug in combine() that also resulted in a bug in read.bsmooth when multiple input directories were specified. Bug reported by Andreas Schoenegger.
Improved combine and fixed a bug. Also added a non-exported combineList for testing.
Bug fix to read.bsmooth; it now works correctly for the default settings (= returning a single object of class BSseq and not a list).
Getting ready for initial release on Bioconductor.
Updated the citations in the vignette(s) and the CITATION file.