Changes in version 1.7.2 BUG FIXES - Fixed cluster assignment issue in function identify_neighborhoods(). Changes in version 1.7.1 BUG FIXES - Added feature_colname parameter to functions plot_cell_marker_levels() and marker_intensity_boxplot(). Changes in version 1.7.0 Development version on Bioconductor 3.20. Changes in version 1.5.3 BUG FIXES - Fixed the wrong layering in plot_cell_categories() and added legend when layer=TRUE Changes in version 1.5.2 NOTES - plot_cell_percentages() now returns the plot. Changes in version 1.5.1 BUG FIXES - No error message when there are duplicates generated by the "TSNE" option of dimensionality_reduction_plot() function. - Deleted duplicated plots returned by plot_cell_categories(). Changes in version 1.5.0 Development version on Bioconductor 3.19. Changes in version 1.3.5 SIGNIFICANT USER-VISIBLE CHANGES - Added a new argument margin_dist for define_structure(). Specifying the margin width with microns instead of layers of cells. Changes in version 1.3.4 BUG FIXES - Added legends to plot_cell_categories() plots when layered = TRUE. - Added a distance parameter to fix the dimension error in calculate_spatial_autocorrelation(). Changes in version 1.3.3 SIGNIFICANT USER-VISIBLE CHANGES - Fixed thresholding bug in predict_phenotypes(). NOTES - Added Shiny App (reading data) link. Changes in version 1.3.2 NOTES - Fixed typo in citation and SPIAT overview diagram. - Moved the following packages from Imports to Suggests: alphahull, plotly. Changes in version 1.3.1 - Added citation to the newly published paper. Changes in version 1.3.0 Development version on Bioconductor 3.18. Changes in version 1.1.6 BUG FIXES - Fixed the mixing score and normalised mixing score calculation. Each reference-reference interaction is now counted once (was treated directional and counted twice) and the fraction of normalised mixing score is fixed. Changes in version 1.1.5 SIGNIFICANT USER-VISIBLE CHANGES - Return message instead of error when there are no cells of interest present in the image (identify_neighborhoods()). - Removed the option of manually defining tumour regions in identify_bordering_cells(). Removed parameters n_of_polygons and draw. NOTES - Moved the following packages from Imports to Suggestions: graphics, umap, Rtsne, rlang, ComplexHeatmap and elsa. SpatialExperiment requires version >= 1.8.0. Removed xROI. Changes in version 1.1.4 BUG FIXES - Fixed error when there are only one cell in the clusters. (identify_neighborhoods()). Changes in version 1.1.3 BUG FIXES - The calculation of cell types of interest to All_cells_in_the_structure in calculate_proportions_of_cells_in_structure() was incorrect. Now fixed. Changes in version 1.1.2 SIGNIFICANT USER-VISIBLE CHANGES - Re-organised the vignettes. Changes in version 1.1.1 BUG FIXES - Fix bug when Cell.ID column is missing from the spe_object in identify_neighborhood(). Changes in version 1.1.0 Development version on Bioconductor 3.17. Changes in version 0.99.14 BUG FIXES - Fixed bugs in identify_neighborhoods(): 1) Assign "Free_cell" to the cells of interest when the number of clustered cells are smaller than min_cluster_size in each cluster; 2) Fixed spe_object output (Adding "Neighborhood" column had a bug previously). Changes in version 0.99.13 BUG FIXES - Minor bug in average_nearest_neighbor_index() - the variable output was not defined when there is no reference cell found in the image. Changes in version 0.99.12 SIGNIFICANT USER-VISIBLE CHANGES - average_nearest_neighbor_index() now returns the index along with the pattern type and the p value. Changes in version 0.99.11 BUG FIXES - Renamed the file of R_BC(). - Fixed a minor bug in identify_bordering_cells() that causes error. Changes in version 0.99.10 SIGNIFICANT USER-VISIBLE CHANGES - mixing_score_summary() ensures returning data in any situation. There will be difference between returning NA and 0. See updated documentation. Changes in version 0.99.9 BUG FIXES - Fixed format_image_to_spe() "general" format. Assigned rownames (markers) and colnames (Cell IDs) to the assay of the formatted spe object. Changes in version 0.99.8 BUG FIXES - Fixed format_image_to_spe() "general" format. The NA intensity value removal was for markers. Instead, it should remove cells that have NA marker intensities. - Fixed predict_phenotypes() plot error. Changes in version 0.99.7 BUG FIXES - Fixed formatting image error when intensity_matrix is NULL under "general" option. Changes in version 0.99.6 SIGNIFICANT USER-VISIBLE CHANGES - identify_bordering_cells() emit Warning when no bordering cells are detected. BUG FIXES - Fixed the message displayed for NA removal when formatting image using "general" format. - Fixed plot functions to display only one plot. plot_marker_level_heatmap(), and plot_distance_heatmap(). Changes in version 0.99.5 SIGNIFICANT USER-VISIBLE CHANGES - Generalised functions in tumour structure families to suit other tissue and cell types, not restricted to tumour and immune cells. The affected functions include: 1. identify_bordering_cells() - deleted "tumour" from the plot title; 2. plot_cell_categories() - cell categories changed to not contain "immune" when feature_colname == "Structure"; 3. calculate_distance_to_tumour_margin() renamed to calculate_distance_to_margin(); 4. R_BT() renamed to R_BC(); 5. calculate_summary_distances_of_cells_to_borders() - one column in the returned data frame has name change; 6. defined_strcture() - parameter name_of_immune_cells renamed to cell_types_of_interest; - identify_bordering_cells() can return the number of clusters of the specified cell type. BUG FIXES - Fixed the calculation of the normalised cross-K AUC in AUC_of_cross_funcion(). - identify_neighborhood() returns an SPE object instead of sending ERROR when the cells of interest do no form any clusters. Changes in version 0.99.4 BUG FIXES - Fixed internal function bind_info() to avoid duplicated columns. Changes in version 0.99.3 BUG FIXES - Added checks for dimensionality_reduction_plot() to return error when the cell count in an image is too low. Changes in version 0.99.2 SIGNIFICANT USER-VISIBLE CHANGES - Added perplexity parameter to dimensionality_reduction_plot(). - Added plot_final_border parameter to identify_bordering_cells(). BUG FIXES - image_splitter() returns NULL for the sub-images that do not contain any cells. - calculate_spatial_autocorrelation() - identify_bordering_cells() only plots the bordering cells when bordering cells are detected and the user chooses to plot them. - The calculation of one cell type to All_cells_in_the_structure in calculate_proportions_of_cells_in_structure() was incorrect. Now fixed. Changes in version 0.99.1 SIGNIFICANT USER-VISIBLE CHANGES - Updated the main object class from SingleCellExperiment to SpatialExperiment. - Replace "sce" with "spe" in function names. - Reading data by format_image_to_spe() using general format is recommended. For other data platforms, format_inform_to_spe(), format_halo_to_spe(), format_codex_to_spe(), and format_cellprofiler_to_spe() are also available. - Deleted "Visium" format option in format_image_to_spe(). - image_splitter() now returns a list of spe objects. Changes in version 0.99.0 Version submitted to Nature Communications. Access the release v0.99.0 here. SIGNIFICANT USER-VISIBLE CHANGES - Systematically renamed parameter names. For example, the parameter for selecting a column of interest is changed from "column" to "feature_colname". - Added a function entropy_gradient_aggregated() for calculating cell colocalisation. - Added a function dimensionality_reduction_plot() for visualisation and quality control. - Added examples to documentation. The example datasets were simulated by spaSim. Changes in version 0.4 SIGNIFICANT USER-VISIBLE CHANGES - Added a new parameter "column" to most functions to select a column of interest. - Users can define cell types based on certain marker combinations or cell properties. - Improved tumour border detection (identify_bordering_cells()). - Added functions to define tumour structure and characterise immune populations in different tissue regions. - Added functions to characterise spatial heterogeneity. - Added functions for cell colocalisation, including mixing score, normalised mixing score, cross K function. Changes in version 0.2 - Version for the paper submitted to BioRxiv.