blisa 1.0.3
- New argument
weighted in plotCCIspatial().
The ligand contribution of each sender cell type at a hotspot is now the
spatially weighted sum over its neighbours.
- New argument
weighted in runCCI(). The
global ligand count of each sender group is now the sum over hotspots of
the local weighted ligand counts used by
plotCCIspatial().
blisa 1.0.2
- Added functionality for visium data
- New function
blisaPathway() for summarizing blisa
results at the pathway level
- Added a new element
level to the blisa
object class, specifying whether the results are lr level or pathway
level
plotLRrank() and plotHotspots() not works
for pathway level results
blisa 1.0.1
- New argument
platform in blisa to handle xenium and
visium
- New argument
as_points in plotHotspots()
to plot dots instead of hexagons
- New argument
size in plotHotspots() to
change the size of the dots
- New argument
background in plotHotspots()
to add background image (e.g., H&E)
blisa 1.0.0
blisa() now uses the fastLISA C/OpenMP
backend by default (fast = TRUE) for the bivariate local
Moran’s I computation. fastLISA moved from Suggests to
Imports. Results are statistically equivalent to the previous
spdep::localmoran_bv backend but not bit-identical
(different RNG).
- New
min_total_counts argument to blisa()
and hexBinCells() drops bins whose total counts (summed
over all genes) fall below the threshold during binning. Default
10.
- New
verbose argument (default FALSE) to
blisa() and hexBinCells() gates progress
messages; blisa() also shows a progress bar over the
ligand-receptor pairs in interactive sessions.
- New
lr_pairs argument to plotCCI() selects
specific ligand-receptor pairs to display, overriding
top_lr.
- New
main argument to plotCCI(),
plotCCILR(), and plotCCIsummary() adds a title
above the heatmap.
blisa 0.2.3
LR_results row names are now informative IDs built from
the ligand/receptor symbols, with subunits joined by | and
the two sides by _ (e.g. TGFB1_TGFBR2|TGFBR1).
The original CellChatDB interaction_name is retained in a
new interaction_name column.
plotCCIspatial() now handles multi-subunit
ligand/receptor complexes (previously errored with
subscript out of bounds).
plotHotspots(), plotCCIspatial(), and
plotCCILR() now match the
ligand/receptor arguments by subunit set,
independent of order, separator (,, _,
|), and whitespace.
blisa 0.2.2
- Added
fast = TRUE option to blisa() which
uses fastLISA::local_moran_bv (C/OpenMP backend) instead of
spdep::localmoran_bv for the bivariate local Moran’s I
computation. Default behaviour unchanged.
- New
cpu_threads argument controlling parallelism when
fast = TRUE.
blisa 0.2.1
blisa 0.2.0