CHANGES IN DMRcate VERSION 1.12.1
- Peak closing sped up with Segment.R

CHANGES IN DMRcate VERSION 1.10.2
- Bugfix for when there are no significant CpG sites at the given threshold

CHANGES IN DMRcate VERSION 1.10.1

- Data filtering for EPIC arrays now incorporates probe information (via DMRcatedata_1.8.3) from Pidsley and Zotenko et al. (2016) Genome Biology 17(1), 208.
- Two new modules are now available in cpg.annotate() in addition to "differential" and "variability": "ANOVA" and "diffVar". "ANOVA" will find whole-experiment DMRs from the entire set of contrasts in the design matrix; "diffVar" finds differentially variable regions (DVMRs) using functionality from the missMethyl package.
- Class GenomicRatioSet (minfi) can now be passed to cpg.annotate().

CHANGES IN DMRcate VERSION 1.8.5

- DMRs can now be called from Illumina's EPIC array. Workflow is identical to that of 450K, just with a different annotation argument to cpg.annotate() and DMR.plot().

CHANGES IN DMRcate VERSION 1.7.2

- Major changes. WGBS pipeline is now implemented with DSS as a regression step instead of limma. 450K pipeline is the same, but with slight cosmetic changes in anticipation of the transition to the EPIC array.
- DMR.plot() has been completely rewritten, now with Gviz and inbuilt transcript annotation for hg19, hg38 and mm10.
- DMRs are now ranked by the Stouffer transformations of the limma- and DSS- derived FDRs of their constituent CpG sites.

CHANGES IN DMRcate VERSION 1.4.1 

- Extra control for Type I error through DMR constituents made commensurate with # of differential limma probes
- CITATIONs added

CHANGES IN DMRcate VERSION 1.0.2

BUG FIXES
•	annotate() renamed to cpg.annotate to avoid clashes with same-named function in ggplot2

NEW FEATURES
•	Kernel estimator has been rewritten from scratch without the need for ks:::kde. Now only takes moderated t-values as cpg weights, as required by the chi-square transformation. 
•	Now allows for multi-level factor experiments, as allowed by limma. Contrasts should be specified with a contrast matrix, otherwise the design matrix MUST have an intercept. (Thanks to Tim Triche Jr. and David Martino for their advice).
•	A GRanges object can be produced from the results.
•	XY probes can also be filtered out using rmSNPandCH().
•	DMR.plot() allows group median lines to be plotted to better visualise distances between groups (Thanks to Susan van Dijk and Magnus Tobiasson for their advice).

