Start of new release series. Version bump reflects changes in dependency packages. See the acidtest, bioverbs, freerange, syntactic, and transformer release notes for more details.
makeSampleData: Switched to S4 method that works on
DataFrameclass objects. Enforcing lower camel case for column names.
makeSingleCellExperiment: Switched to S4 method approach that currently requires
SimpleListinput for assays. Previously, we allowed
listinput for assays, but we’re tightening this up to simply the package code base and improve consistency.
aggregatemethods now consistently return the primary assay named as
counts. This follows the recommended conventions defined in SingleCellExperiment. Aggregation functions will now intentionally fail for
SummarizedExperimentobjects that don’t contain an assay named
calculateMetrics: Migrated code here from bcbioSingleCell. Improved method to support
DelayedArrayclass for large matrices.
methodFormalscalls to have better backward compatibility with R 3.5.
calculateMetrics: Bug fix for getting
use.names = TRUE. This improves Bioconductor backward compatibility. Also updated internal code to call
readSampleDatainternal code, but still supporting bcbio pipeline conventions (i.e. “description” column for samples) as the default. I’ve reworked this approach so we can also call
readSampleDatainside the Chromium package (for 10X Genomics single-cell RNA-seq data) without having to depend on the bcbio R packages.
aggregateCols: Sped up return by calling
makeSingleCellExperiment. Note that this now doesn’t return session information in the object.
cell2sample: Renamed return from
tbl_df, for consistency.
assert*functions that have been reworked using a goalie approach.
estimateSizeFactors: Removed option to calculate “deseq2-median-ratio” using DESeq2. We may revisit this idea in a future release.
makeSampleData: Improved internal code for
selectfrom AnnotationDbi, now that we’re no longer depending on dplyr.
relevelon S4 objects internally, where applicable. This is better supported by S4Vectors package.
meltCounts: Added initial method support for
SingleCellExperiment. Currently requires deparsing of the count matrix to call
reshape2::melt. Returns columns with S4 run-length encoding (Rle) applied.
mapGenes: Converted warning to message when
strict = FALSE.
makeSummarizedExperiment: Now automatically handles non-Ensembl gene symbols present in assays that aren’t defined in
rowRanges. This applies primarily to 10X Cell Ranger v3 output, which includes some non-Ensembl gene symbols: CD11b, CD4, CD8a, HLA-DR, IgG1, PD-1, etc. The function still intentionally errors on unannotated Ensembl features, which often indicates an accidental release version mismatch.
Updated R dependency to 3.6.
melt: Added S4 methods for melting data into long format. Provides support for
nonzeroRowsAndCols: Quickly remove non-zero rows and columns from a matrix or
calculateMetrics: Now calls
nonzeroRowsAndColsinternally first when
prefilter = FALSE, speeding up calculations significantly for very large
SingleCellExperimentobjects. This was added to improve loading of example unfiltered 10X Genomics Chromium data.
autopadZeros: Added improved support for detection and automatic handling of zeros in need of padding on the left side of a character vector. This addition is necessary for handling of Genewiz processed FASTQ file names.
makeTx2Genefunctions now support
ignoreTxVersionargument, similar to conventions defined in tximport package.
minargument now defaults to
-Inf, and the
minMethodnow defaults to
"absolute", instead of
"perRow", since this behavior is more intuitive to the user.
makeSampleData: Made function slightly more flexible. Now allowing automatic rowname coercion from columns (“sampleID”, “rowname”, “rn”), similar to approach employed by data.table and tibble packages.
stripGeneVersionsalias, which uses the same code internally as
is_pristinebug in DelayedArray v0.11.8 is fixed on Bioconductor Devel (3.10). This is causing unit tests to fail otherwise. See related issue
importSampleData: Pipeline now defaulting to “none” instead of “bcbio”, since this flag is now properly hard coded in bcbio R packages. Added new Constellation (CPI) pipeline option.
makeSampleData: Now checks for all NA columns and rows, similar to approach in
importSampleData. This helps improve return consistency. Automatic rowname setting has been tweaked a bit to no longer attempt to remove original ID column.
wormbaseVersion. Similar shell variants are available in the koopa package.
correlation: Added S4 method support that mimics base
cormethods, but is more flexible, supporting additional arguments via
...in generic. This way we can provide intelligent and quick correlation calculations for nested assays inside a
DESeqResults(see DESeqAnalysis package).
formalsListglobal slightly. Using the
synesthesiacolor palette for
heatmap.colorargument doesn’t always perform well enough, so I’m switching to a blue/black/yellow palette defined by
blueYellowin acidplots instead for
synesthesiapalette performs really well for correlation heatmaps, and is now recommended by default via the
acid.heatmap.correlation.colorglobal now instead.
filterCells: Improved internal sampleName / sampleID handling.
integerCounts: Simple method support for returning a rounded integer counts matrix. Intended primarily for downstream handoff to bulk RNA-seq differential expression callers, such as DESeq2.