Solved issue with usage of the torch
library
Allowed user to explicitly opt-out from the torch
library usage:
COTAN will avoid torch
commands when the option "COTAN.UseTorch"
is
set to FALSE
Added support for the torch
library to help with the heavy lifting
calculations of the genes' COEX matrix, with consequent substantial speed-up,
especially when a GPU is available on the system
First release in Bioconductor 3.20
Refactored DEAOnCluster()
to make it run faster.
Now clustering functions dump the GDI
check results for all clusters
Changed default GDI
threshold to 1.43
Added new input to mergeUniformCellsClusters()
to allow proper resume of
interrupted merges
Added possibility to query whether the COEX
matrix is available in a COTAN
object
Made checks more strict when adding a clusterization or condition
Increased reliability of clustering functions by improved error handling and by allowing retry runs on estimators functions
Speed-up of GDI calculation via Rfast package
Added possibility of using distance between clusters based on Zero-One matrix instead of DEA
Added average floor to logFoldChangeOnClusters()
to dampen extreme results
when genes are essentially absent from a cluster.
Added method to handle expression levels' change via log-normalized data:
logFoldChangeOnClusters()
Minor fix in the import of operators to align to new version of roxigen2
Restored default adjustment method of pValueFromDEA()
to "none"
for backward compatibility reasons
Solved issue with cleanPlots()
when the number of cells exceeded 65536
Added methods to calculate the COEX
matrix only on a subset of the columns
Now the function pValueFromDEA()
returns the p-value adjusted for multi-test
Stopped using explicit PCA via irlba package: using BioConductor PCAtools::pca instead
First release in Bioconductor 3.19
Made passing clusterizations to COTAN
functions more easy:
now all functions that take a COTAN
object and a clusterization as input
parameters can also take a clusterization name
Added time-stamps to log entries when written on a log file
Fixed bug in the clustersMarkersHeatmapPlot
function when given a
clusterization not matching the latest added to the COTAN
object
Fixed issue with the highest possible resolution in seuratClustering()
function, needed when large datasets must be split in many clusters
Added new flag to the function cleanPlots()
to suppress evaluation of the
PCA on the normalized data. In particular, this allows to reduce significantly
time spent within the function checkClusterUniformity()
Added initialResolution
parameter to cellsUniformClustering()
: it allows
users to specify the initial resolution used in the calls to
Seurat::FindClusters()
method. It now uses the same default as Seurat
Added new method estimateNuLinearByCluster()
that calculates nu
ensuring
that its average is 1.0 in each given cluster
Added function reorderClusterization()
: it reorders the given clusterization
so that near clusters have also near labels
The functions cellsUniformClustering()
and mergeUniformCellsClusters()
now
return the result of this new function
Separated p-value calculations from DEAOnClusters()
into the new function
pValueFromDEA()
. Those data.frames
are no longer part of the list returned
by the functions DEAOnClusters()
and mergeUniformCellsClusters()
Added function getClusters()
to retrieve the wanted clusterization from the
cells' meta-dataset
Added function calculateGenesCE()
: it returns the cross-entropy between
the expected absence of a gene reading against the observed state
Fixed minor issue with logThis()
to file: it was always appending a new line
even when appendLF
was set to FALSE
Now checkClusterUniformity()
returns more GDI stats like the percentage of
genes above threshold or the last percentile of the GDI values
Revamped mergeUniformCellsClusters()
to select in order all the the most
likely candidates pairs of clusters to merge. Provided new user parameter to
balance the merging of most possible candidates versus the time spent doing so
Improved dropGenesCells()
method: it now retains all meta-data information
that is not related to the results of the other methods
Added zoomed UDE plot to cleanPlots()
return. It suggests a possible
cut level for low UDE cells
Improved mergeUniformCellsClusters()
: now it attempts to merge more
clusters pairs
Now errors in the seuratClustering()
function are interpreted as remaining
cells not-clustered "-1". This applies mostly to cases when Seurat
finds only
singlets
Added flag calcCoex
to proceedToCoex()
and automaticCOTANObjectCreation()
functions to allow user not to spend time calculating the genes' COEX when not
needed
Solved potential issue in the clustersMarkersHeatmapPlot()
regarding clusters'
labels
Added new internal function niceFactorLevels()
that ensures all the factors'
levels will have labels with the same length, via padding the integers values
with '0' and string values with '_'
Relaxed tolerance on tests comparing against saved data
Speed-up by use of parallelDist::parDist()
to calculate distances instead of
stats::dist()
Fixed regression tests failing on non-Linux architectures
Completed function clustersMarkersHeatmapPlot()
Added new utility function normalizeNameAndLabels()
Added mergeClusters()
and multiMergeClusters()
functions
Added support to conditions
in cells' meta-data
Now clusterizations are stored as factors
Fixed COTAN::validity
method in AllClasses.R
Fixed bug in proceedToCoex()
in cases when saveObj == TRUE
Renamed methods dealing with housekeeping genes and fully-expressed cells to use the more proper names fully-expressed genes and fully-expressing cells
Added possibility to users to set the cutoff and thresholds used by the clean
and related methods
First release in Bioconductor 3.18
Solved remaining documentation warnings
Updated the vignette, README.md
and NEWS.md
Dropped second vignette: will be merged in the other one...
Minor bug fixes and new function clustersMarkersHeatmapPlot()
Included new functionalities for Bioc 2.17 release:
created a new COTAN
class to replace the old scCOTAN
: this class provides
internal invariants along a wide host of accessors that allows users to avoid
peeking inside the class
made a new multi-core implementation of the model parameters estimations and
COEX
calculations that achieves much higher speeds.
added new functionality about gene clusters starting from given markers lists
added new functionality about uniform cell clustering based on the maximum GDI level in the cluster
added function to get a differential expression estimation for each cluster against background
added function to get an enrichment score for each cluster given a list of markers specific for the cells' population
added plots to asses data-set information at cleaning stage
After official release. PCA function changed to avoid basilisk and Python.
Release before the official release of Bioc 3.15. Main changes: - The way in which the COEX matrix is estimated and stored is changed to occupy less space and run faster.
Initial Bioconductor release
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.