- Moved some code to the new package
sccore
- Fixed inconsistent use of parameters for different spaces in
buildGraph
- Various small fixes
- Fixed the number of components calculated for the simple PCA rotation
- Conos is R6 class now (instead of refClass)
- Functionality for PAGA graph collapsing
- Parameters
k.same.factor
andbalancing.factor.per.sample
tobuildGraph
. It can be used to improve alignment between different conditions: withsame.factor.downweight
it gives the system similar tok.self
andk.self.weight
- plotDEheatmap() function for viewing marker genes
- Function
convertToPagoda2
to create Pagoda 2 from Conos. Helpful for PagodaWebApp.
- Fixed
getDifferentialGenes
- Fixed testing clustering stability
- Added mean M value column to the diff. expression output
- Optimized plotting with coloring by genes
getDifferentialGenes
uses first clustering by default- Fixed bug with
collapseCellsByType
. Note: probably will affect DE results. - Added re-normalization of edge weights to fix problem with negative edge weights during label propagation
- Now in plotting 'groups' aren't ignoted if 'gene' is provided: it's used to subset cells for plotting.
- UMAP now set
n_sgd_threads
fromuwot
ton.cores
by default. It gives much better parallelization, but kills reproducibility. Usen.sgd.cores=1
to get reproducible embeddings. - Account for
target.dims
in UMAP embedding - Fixed estimation of
cor.based
withalingnment.strength == 0
. It removes edges with negative correlation and reduce down-weight of inter-sample edges, which can change results of the alignment. - Changed default value of
fixed.initial.labels
inpropagateLabels
fromFALSE
toTRUE
. Presumably,FALSE
should never be used. - New output format for label propagation (list of "labels", "uncertainty" and "label.distribution")
- Numerous small bug fixes and small validations for correct arguments
- ScanPy integration tutorials to refelect the changes in
saveConosForScanPy
- Metrics to masure specifisity of cell type markers to DE info in
getDifferentialGenes
(parametersappend.specifisity.metrics
andappend.auc
) - Implementation of label propagateion based on matrix equations (occured to be too slow)
- Function
findSubcommunities
to increase resolution for specific clusters - Parameter
subgroups
toembeddingPlot
. It allows to plot only cells, belonging to the specified subgroups - Parameter
keep.limits
toembeddingPlot
- Added metrics to masure specifisity of cell type markers to DE info in
getDifferentialGenes
(parametersappend.specifisity.metrics
andappend.auc
) velocityInfoConos
function for RNA velocity analysis on samples integrated with conos (together with supplementary functionsprepareVelocity
andpcaFromConos
)- "Running RNA velocity on a conos object" section in README.md (explains usage of the
velocityInfoConos
function) - Function
getJointCountMatrix
to conos obect - New possibilities to customize the output of
saveConosForScanPy
- Function
parseCellGroups
to parse properly cell groupings depending on user settings - Function
estimteWeightEntropyPerCell
to visualize alignment quality per cell
- Changed a description line in the getClusterPrivacy() doc to fix installation under some R3.6 versions (issue 32)
- Fixed docker build: use of BiocManager, reference to master instead of dev
- Updated src/Makevars to remove the CXX directive, which trips up older versions of R (3.2.x)
- Support for CCA space
buildGraph
now use PCA space as the default- fixed common variance rescaling to use geometric mean of the target
- Support for Seurat v3 objects
- Functions to export Conos object to ScanPy
- Default value for
cluster.sep.chr
in DE functions is changed from '+' to '<!!>', as it shouldn't be normally present in cluster names - Removed Boost dependency
- Fixed version of Seurat and fpc packages in Docker
- Fixed ggplot2 namespace for function calls in
plotClusterStability
- Renamed
stable.tree.clusters
,get.cluster.graph
andscan.k.modularity
- Removed exports of largeVis internals,
get.cluster.graph
andget_nearest_neighbors
embedding
is stored with samples by rows now (i.e. not transposed anymore)- Using scale.data instead of data in Seurat if provided
- multitrap.community and multimulti.community functions
- Pre-release version