Changes in version 1.28.0                        

    o   Use a more stable algorithm for identifying the knee point in
	barcodeRanks(). The new algorithm is based on maximizing the
	distance from a line between the plateau and the inflection
	point. Previously, we tried to minimize the signed curvature
	but this was susceptible to many local minima due to the
	instability of the empirical second derivative, even after
	smoothing.

    o   Set alpha=Inf as the default for testEmptyDrops(). This is
	motivated by the realization that an underestimated alpha can
	still yield anticonservative p-values and is not universally
	safer than alpha=Inf. Defaulting alpha=Inf is preferable as it
	is at least correct in the expected case of multinomial
	sampling.

                       Changes in version 1.18.0                        

    o   Added an intersect.genes= option to read10xCounts() for samples
	with inconsistent gene information. Automatically fix empty
	chromosome names for mitochondrial genes in certain Cellranger
	outputs.

                       Changes in version 1.12.0                        

    o   Added BPPARAM= to read10xCounts() for parallelized reading of
	multiple samples.

    o   Gave all the *Ambience() functions better names, and
	soft-deprecated the current versions.

    o   Added ambientContribSparse() to estimate the ambient
	contribution under sparsity assumptions.

    o   Added cleanTagCounts() to remove undesirable barcodes from tag
	count matrices.

    o   Converted all matrix-accepting functions to S4 generics to
	support SummarizedExperiment inputs.

    o   emptyDrops() will now coerce all DelayedArray inputs into
	wrapped SparseArraySeeds.

    o   Setting test.ambient=TRUE in emptyDrops() will no longer alter
	the FDRs compared to test.ambient=FALSE. Added test.ambient=NA
	to retain back-compatible behavior.

    o   Bugfix for correct use of redefined lower when by.rank= is set
	in emptyDrops().

    o   Added a constant.ambient=TRUE option to hashedDrops() to better
	support experiments with very few HTOs.

                       Changes in version 1.10.0                        

    o   Migrated downsampleMatrix() to scuttle with a re-export.

    o   Added features= to downsampleReads() for per-feature-set
	downsampling.

    o   Added matrix support for y= and ambient= in maximumAmbience().

    o   Added controlAmbience() for easy estimation of ambient
	contamination with control features.

    o   Added removeAmbience() function to remove the ambient solution
	from a count matrix, mostly for aesthetics.

    o   Report library index and feature type in output of
	read10xMolInfo().

    o   Support subsetting by library index/type in functions that use
	the molecule information file, such as swappedDrops() and
	chimericDrops().

    o   Added by.rank= option to estimateAmbience() and emptyDrops(),
	for estimation of the ambient profile by excluding barcodes
	with the largest totals.

    o   Added exclude.from= option to barcodeRanks(), to avoid problems
	with instability at low ranks for knee/inflection calculations
	(contributed by Stefano Mangiola).

    o   Minor bugfix in barcodeRanks() calculation of the knee point.
	Note that this affects the default choice of retain= in
	emptyDrops().

    o   Split off HTO ambience inferences into a separate
	inferAmbience() function.

    o   Added support for combinatorial barcodes in hashedDrops().

                        Changes in version 1.8.0                        

    o   Added the downsampleBatches() function for convenient
	downsampling of batches.

    o   Preliminary support for using the output of write10xCounts()
	back in Cellranger.

    o   Support reading in 10X output files via prefixes in
	read10xCounts(). Automatic detection of whether a file is
	Gzipped or not.

    o   Added chimericDrops() to remove chimeric molecules due to
	within-sample re-priming.

    o   Added hashedDrops() to demultiplex cell hashing experiments.

    o   Added maximumAmbience() to estimate the maximum contribution of
	the ambient profile.

                        Changes in version 1.6.0                        

    o   Switched emptyDrops() to use Boost's discrete_distribution for
	weighted sampling.  This results in some minor stochastic
	changes to the Monte Carlo p-values. Automatically round
	non-integer count matrices.

                        Changes in version 1.4.0                        

    o   Removed read10xMatrix().

    o   Supported CellRanger v3 output files in read10xMolInfo(),
	read10xCounts(), write10xCounts().

    o   Modified barcodeRanks() to return a DataFrame with
	knee/inflection estimates in metadata.

    o   Slight change to random number generation in emptyDrops() to be
	agnostic to number of cores.

                        Changes in version 1.2.0                        

    o   Added removeSwappedDrops() for removing swapping in other types
	of droplet-based data.

    o   Added alpha= argument to testEmptyDrops() to support
	overdispersion during sampling. Returned arguments and
	estimates in metadata of testEmptyDrops(), emptyDrops().

    o   Added encodeSequences() for convenient 2-bit encoding of
	sequences.

    o   Added get10xMolInfoStats() function to compute per-cell
	statistics from a molecule info file.

    o   Deprecated read10xMatrix(), as it does not add much practical
	value over Matrix::readMM().

    o   Support the 10X sparse HDF5 format in read10xCounts().

    o   Support the 10X sparse HDF5 format in write10xCounts().

                        Changes in version 1.0.0                        

    o   
	New package DropletUtils, for handling droplet-based
	single-cell RNA sequencing data.