- Fixed wrong parameter interpretation in
neighborhood_count_matrix()
- Fixed incompatibility with
julia-1.9
(#78) - Fixed
boundary_polygons_from_grid
error (#82) - Fixed a problem with huge integers in cell ids (#83)
- Fixed a bug with prior segmentation loading
- Fixed a bug with compartment genes
- New output cell QC parameters
avg_assignment_confidence
,max_cluster_frac
andlifespan
- Segmented cells are now saved to loom instead of TSV. To return an old behavior, use
count-matrix-format="tsv"
- Minimal multi-threading (see README)
iters
andn-cells-init
parameters were removed from the CLI shortcuts. To change them, use the config or--config.segmentation.iters
and--config.segmentation.n_cells_init
parameters (see 'Advanced configuration section in the readme').
- Breaking changes in config file structure and CLI
- Greatly improved responsiveness of the CLI and simplified installation process
- Major refactoring of the code
- Various performance improvements
- Faster and more precise algorithm for estimating boundary polygons. Now each cell has exactly one polygon in the output GeoJSON.
- For method details see Awrangjeb, 2015, it's pretty similar.
- Closes #15 and #41, potentially also #32 and #37.
- Using sparse PCA for NCV estimation on large datasets
baysor segfree
output is now fully compatible with loom v3 format- Cells and NCVs now have IDs in the format
{type}{run_id}-{cell_id}
, wheretype
isC
for cells andV
for NCVs, andrun_id
is a unique ID of Baysor run --save-polygons
now works regardless of-p
- Fixed some package versions, dependencies should cause fewer bugs now
- Fixed some bugs
- Fixed random seed for all CLI runs
- Adjusted some CLI parameters
- Added
segfree
run option to extract NCV vectors
- Slightly optimized compilation time
- Minor updates of core formulas
- CLI parameter
no-ncv-estimation
to disable estimation of NCVs
scale-std
now can be specified from CLI parameters- Several bugs fixed
- Allow missing genes in the input data
- Updates in the core algorithm
- All diagnostic plots were updated
exclude-genes
option that removes genes from the data frame before segmentation.- Segmentation of compartments based on the list of compartment-specific genes
- Using information about compartment per molecule in the segmentation algorithm when available
- 3D segmentation
- The data can not be split by frames anymore. This functionality didn't work well previously and was hard to maintan.
- Fixed the CLI installation bug
- More information in logging
- Better initialization for molecule clustering. This should improve cell separability a lot!
- Some memory optimization
- Fixed plotting performance by moving to Makie.jl
- Estimating scale when prior segmentation is provided as a CSV column
- Added the option
--save-polygons=GeoJSON
to save cell boundary polygons in the GeoJSON format
- Fixed Makefile julia version
- Improved polygon visualization
- Regressed to Plots 1.6.0 because of the performance issues
- Fixed docker build
- Small bug fixes
min-molecules-per-segment
parameter is working now- Fixed visualization of prior segmentation
- Added travis config
- Saving NCV colors to the
ncv_color
field of segmentation.csv
- Dropped support for julia < 1.5
- Updated dependencies
find_grid_point_labels_kde
now preserves label ids- Fixed docker build
- Added ImageMagick dependency to fix problems with DAPI prior