NEWS.md
map_celltypes_sce(). The subsampling is now done per cluster to avoid smaller clusters being excluded.reduce_dims_sce() function to use the latest monocle3 version and to be able to pass residual_mod_formula_str to regress out variables.map_celltypes_sce() function to generalise the celltype annotation.Marker identification for groups of cells. The find_marker_genes() function identifies marker genes for arbitrary groups of cells (e.g. celltypes, clusters). The function is now invoked by the annotate_celltype_metrics() function to generate marker genes for clusters and celltypes which are presented in the updated report_celltype_metrics() function.
Adaptive thresholding on mitochondrial counts with now enabled with annotate_sce() and reflected in the associated report report_qc_sce().
Adaptive thresholding for per-sample QC. The annotate_sce() function now accepts “adaptive” as a threshold value for upper limits on various cell metrics together with a Median Absolute Deviation (MAD) threshold (default 3.5). This allows QC thresholds to be determined on a per-sample basis, improving integration of datasets from different sources. The report_qc_sce() function now plots and describes these adaptive thresholds.
Dirichlet modeling for statistically significant differences in cell-type abundance. The model_celltype_freqs() function allows flexible cell-type differential abundance analyses, with an accompanying function to generate a report report_celltype_model().
Cell type metric annotations generated by annotate_celltype_metrics() with an accompanying report produced by report_celltype_metrics(). Currently includes three sections: (1) reduced dimensionality plots, (2) cell-type proportions by groups, and (3) distributions of various metrics for each cell-type.
Differential gene expression reports with report_de() using the result table generated by the perform_de() function after differential expression analysis. The report includes all the parameters used during the differential expression analysis, a summary of the up and downregulated genes, a volcano plot showing top 10 up and down regulated genes, and the result table.
Dataset integration performance analyses (including kBet) and plots produced by annotate_integrated_sce() with an accompanying report produced by report_integrated_sce():
Finding outlier datasets (individuals): LIGER takes the union of variable genes across datasets and use them for integration. Therefore, it is important to find outlier datasets. Now integrate_sce() and liger_preprocess() functions have been modified in order to generate a list of variable genes for each of the datasets. The new annotate_integrated_sce() function generates Venn and Upset plots using this data in order to visualise sizes of isolated dataset participation to the total variable genes used for integration. This helps to identify outlier dataset(s).
Batch effect correction by LIGER: now annotate_integrated_sce() provides proofs for batch effect correction by LIGER. This function quantifies the batch effect caused by each of the categorical covariates for data generated by PCA vs LIGER. The kBET method is used for quantification of batch effects, and results are generated as rejection rate box plots as well as kBET test P-values. For each of the comparisons annotate_integrated_sce() also visualises the batch effects using tSNE plots. annotate_integrated_sce() now generates UMAP plots to visualise clusters identified using PCA vs LIGER data.
Interactive report for dataset integration, dimension reduction, and clustering: The new report_integrated_sce() function generates an interactive report which includes method summary, key parameters used, Venn and Upset plots for variable genes used for integration, PCA vs LIGER side by side comparison of batch effect quantified by kBET and visualised by tSNE, and UMAP plots showing clusters identified using PCA vs LIGER data.
Improvements in the QC report generated by report_qc_sce(), including new CSS styling. Key information is now summarized at the beginning of the report.
Pseudobulking algorithm improved to utilize matrix multiplication instead of lapply: typical 1-2 orders of magnitude speed increase.
annotate_merged_sce() and report_merged_sce() to examine QC metrics across samples to facilitate identification of problematic samples.
find_cells() annotates cells / empty drops using the EmptyDrops algorithm. The single-sample QC report generated by report_qc_sce() now includes EmptyDrops results with metrics and plots of algorithm performance.
integrate_sce() runs LIGER on a SingleCellExperiment created by merge_sce(), storing the H.norm factors in a reducedDim() slot available as an input for dimensionality reduction.
reduce_dims_sce() now includes a parameter for input_reduced_dim which accepts one or more of PCA and Liger as input, producing tSNE_PCA, tSNE_Liger, UMAP_PCA, UMAP_Liger etc. allowing dimensionality reductions with and without integration to be examined.
Improved annotation of SingleCellExperiment celltypes by map_celltypes_sce() to enable three new functions to write (write_celltype_mappings()), read (read_celltype_mappings()), and update (map_custom_celltypes()) the celltype mappings for a SingleCellExperiment. This enables cluster -> celltype mappings to be quickly and easily revised if needed.
find_impacted_pathways() and report_impacted_pathways() allow a differential gene expression table to be submitted to WebGestalt and ROntoTools (and in future additional tools) and an interactive report to be generated.