Important If you have objects created since 2.0.0 but with a version < 2.3.0 (i.e. including 2.0.0), you should run updateObject
to update the class definition because there have been changes to the class definition since that version:
ceObj<-updateObject(ceObj)
Warning This command will, however, loose information saved about the last mergeClusters
call that you have made if your object is from version < 2.1.4. You may want to save that information and manually update the slots. If you do so, make sure you call validObject
to make sure that you have done so correctly (in particular, you will have to have a value for the slot merge_demethod
, see ?ClusterExperiment
which is a new slot). For example,
ceObjNew<-updateObject(ceObj)
ceObjNew@merge_index<-ceObj@merge_index
<etc>
If you have objects from before 2.0.0 (when the class was called 'clusterExperiment'), you should construct a new object using the ClusterExperiment
function. For example,
ceObjNew<-ClusterExperiment(
se=as(ceObj,"SingleCellExperiment"),
clusterMatrix=ceObj@clusterMatrix,
<etc>
)
See ?ClusterExperiment
for the names of the slots.
There have also been a number of changes and enhancements to the package. These are the most important (a complete list is detailed in the NEWS file of the package -- all releases since May 1, 2018)
- We have changed the function
combineMany
tomakeConsensus
. This has resulted in changes to the names of the arguments ofRSEC
combineProportion
->consensusProportion
inRSEC
combineMinSize
->consensusMinSize
inRSEC
- Add functionality to
getBestFeatures
to allowedgeR
for DE, as well as weights used withedgeR
for compatability with weights to handle zero-inflation. As part of this changeisCount
argument has been replaced with more fine-grainedDEMethod
argument ingetBestFeatures
,mergeClusters
; and the argumentmergeDEMethod
inRSEC
is now available. - We have changed the argument
sampleData
in various plotting commands tocolData
to better indicate that the argument is to identify columns incolData
that should also be plotted. FurthermoreplotDendrogram
now takes the argumentcolData
for plotting of information incolData
with the dendrogram. - We have changed the names of arguments related to unassigned (
-1
or-2
assignments) to more consistently use the term "unassigned", as well as adding the functionassignUnassigned
:- argument
removeNegative
->removeUnassigned
ingetBestFeatures
- argument
ignoreUnassignedVar
->filterIgnoresUnassigned
inmergeClusters
(and other functions) for clarity. - function
removeUnclustered
->removeUnassigned
- argument
- New plotting functions:
plotTableClusters
plotFeatureScatter
- Allow the arguments
subsample
andsequential
toRSEC
to allow for opting out of those options for large datasets (but default isTRUE
unlikeclusterMany
) - The argument
whichAssay
is added to most functions to allow the user to select the assay on which the operations will be performed. - We've changed how we store the cluster hierarchies so that we now use the
phylo4d
class ofphylobase
package (previously we stored them as adendrogram
class). This makes it easier to store information about the dendrograms and manipulate them. There are various helper functions related to this change. See?clusterDendrogram
. - We now store the coClustering matrix in the
coClustering
slot as asparseMatrix
class from the packageMatrix
. This will reduce the size of the object in memory.