.read_matrix_block()dreamlet() gives clearer error message for singular design matrixplotGeneHeatmap() handles zmax correctly nowprocessOneAssay, set rescaleWeightsAfter=FALSE by defaulttopTable() with multiple coef valuesrun_mash() with multiple coefficients
dreamlet() allow formula to include only interceptcompositePosteriorTest()plotPCA() and outlierByAssay()
list, not just dreamletProcessedDatapbWeights()outlierByAssay()eBayes()processAssays() pass argument scaledByLib to voomWithDreamWeights()pbWeights()getExprGeneNames()seeErrors() and documentationoutlier() to compute z-scores. How returns data.frame()outlierByAssay() and plotPCA()compositePosteriorTest() allows exclude set to be NULLmeta_analysis()stackAssays() now includeds metadata()$aggr_means correctlycompositePosteriorTest()get_metadata_aggr_means() to extract aggr_means when SCE is produced by cbind'ingaggr_means in aggregateToPseudoBulk()rdf is low for all genesplotBeeswarm()rowWeightedVarsMatrix()isFullRank() check in dreamlet()run_mash()pbWeights() add argument maxRatiopbWeights()dreamlet()pbWeights() to compute precision weights for pseudobulk countsextractData() and include it in vignettestackAssays()diffVar()getVarFromCounts) so zeta is a mean, not a sumcomputeLogCPM() uses augmentPriorCount()computeLogCPM() now returns matrix instead of sparseMatrixvariancePartition version dependencygetWeightsList()processAssays()setAutoBlockSize() update within aggregateToPseudoBulk()processAssays() and fitVarPart()styler::style_pkg()dreamletCompareClusters() now allows cell-level covariates in response to https://github.com/GabrielHoffman/dreamlet/issues/11dreamlet::residuals()processAssays() use voomWithDreamWeights(..., span="auto") to estimate the lowess tuning parametermerge_metadata() when a cell type is not observed for all donors.dreamlet() fix issue when contrasts are specified and formula includes variable from metadata()assays argument to buildClusterTreeFromPB()processAssays() when assays is droppedzellkonverter (>= 1.10.1) to avoid issues with previous version
topTable() for dreamletResult in the case where one or more cells didn't estimate the coefficient of interestcomputeNormCounts() and computeLogCPM()topTable() to deal with multiple coef as arraycolData<- for dreamletProcessedDatatopTable() and plotForest()aggregateToPseudoBulk() stores mean of cell-level covariates in metadata(pb)$aggr_meansprocessAssays(), dreamlet(), fitVarPart()aggregateNonCountSignal()plotProjection()outlier()plotForest()plotVolcano() to allow scales="free_y"
aggregateNonCountSignal() to include filtersaggregateNonCountSignal()aggregateNonCountSignal()plotGeneHeatmap() drop empty genesbuildClusterTreeFromPB()topTable()as.dreamletResult()variancePartition dependency and sourceprocessAssays() and processOneAssay(), add argument min.prop indicating the minimum proportion of retained samples with non-zero countszenith_gsa() for few gene setscomputeCellCounts()transpose argument to plotGeneHeatmap()alpha arugment to plotVoom()plotVarPart()totalCPM column to output of cellTypeSpecificity() to use for filtering. Functions dreamlet::plotHeatmap() plotViolin() and plotPercentBars() now ignore this columnplotGeneHeatmap()
assays to plotVarPart()extractData()aggregateToPseudoBulk() by speeding up check in .check_arg_assay()
tabToMatrix()topTable() when all random effects are droppedaggregateToPseudoBulk() when summarizing for just 1 samplegetTreat() for dreamlet() resultdroplevels for colData in processAssays()processAssays() to detect issues with SCEaggregateToPseudoBulk() with sample orderingdreamletCompareClusters()colsum2() using beachmat code.aggregateToPseudoBulk() by fixing `aggregateByColnames()run_mash() to combine results across coefsdreamlet::colsum_fast() used in pseudobulkda_to_sparseMatrix()aggregateToPseudoBulk() for DelayedArray now uses colsum_fast()
DelayedArraydreamletCompareClusters():
plotZenithResults()errorsAsWarnings. If TRUE warns and returns NULL.dreamletCompareClusters() to be compatible with zenith_gsa()
formula in dreamletCompareClusters()aggregateToPseudoBulk() when a Seurat object is used
aggregateToPseudoBulk() when a Seurat object is used
aggregateToPseudoBulk() when a Seurat object is usedcollapse=TRUE to dreamletCompareClusters()dreamletCompareClusters()dreamletCompareClusters()min.samples to processAssays(), processOneAssay()dreamletCompareClusters() and run_mash()dreamletCompareClusters()mashr dependencyrun_mash()
zenith_gsa(), plotVolcano(), plotForest() for resultscellTypeSpecificity() for genes with zero reads across all cell typesplotForest() and zenith_gsa() changed for consistancyremoveConstantTerms() when excluded variable string (i.e. tissue) is also a substring of other variables (i.e. tissueStatus)residuals() for dreamlet() resultdreamletPairs()removeConstantTerms() with multiple constant termscellTypeSpecificity() by adding plotPercentBars() and plotViolin() compatabilitytopTable() when coef is not estimatedassays to dreamlet(), fitVarPart(), and processAssays()processOneAssay() weights by number of cellsvariancePartition >= 1.25.1 to handle weights in voomWithDreamWeights()topTable()plotPercentBars() for class vpDFapplyQualityWeights()ilr_composition_test.RdreamletResult using coefNames()regModel()removeConstantTerms() now doesn't drop terms solely because of NA's
Oct 5, 2021
suppress package startup messages in aggregateToPseudoBulk()
bug fix in removeConstantTerms()
Sept 30, 2021
call to zenith_gsa() adds argument inter.gene.cor and progressbar
fix output to cellTypeCompositionVarPart() and cellTypeCompositionTest()
fix issue with topTable() where FDR was evaluated on only a subset of genes
dreamlet() to handle linear contrastsremoveConstantTerms() now drops categorical variables with only a max of one example per categorycellTypeCompositionTest()
processAssays() to include extra meta-data