- Added
aggregation,center, andscaleparameters torun_pca()function, allowing users to tune PCA computation for single-cell and/or pseudobulk data.
- Added comprehensive test coverage for the
aggregationparameter inrun_pca(), including selective aggregation and error handling. - Enhanced README documentation with link to external resources page (tutorials, case studies, and metric explanations).
- Fixed incorrect name assignment in
run_pca()when using selectiveaggregationparameter—now correctly assigns names based on computed aggregations rather than all data slots.
- Updated both vignettes to use fully runnable toy data workflows for matrix, Seurat, and SingleCellExperiment inputs.
- Removed non-justified
eval = FALSEvignette chunks and aligned examples with the current API. - Updated reciprocal classification vignette examples to use
reduction = FALSE, avoiding unsupported dimensional-reduction usage forrecip_classifmethods. - Standardized function and parameter naming to snake_case across the active package API.
- Curated naming/style consistency in active source, including replacement of remaining
1:...indexing patterns.
- Removed
get.optimal_clusteringfrom the public package workflow/documentation while it remains in development.
- Initial Bioconductor submission with ground-truth-agnostic cell type evaluation framework
- Support for multiple input formats: Seurat, SingleCellExperiment, and count matrices
- Multiple dissimilarity methods:
- Pseudobulk-based distances (Euclidean, Cosine, Pearson)
- Wasserstein distance on single-cell distributions
- Reciprocal classification approaches
- Comprehensive internal validation metrics:
- Silhouette scores
- Neighborhood Purity
- Ward's clustering consistency
- Orbital medoid distances
- Average similarity measures
- 2-label silhouette analysis
- Gene selection methods:
- Highly variable genes (HVG) detection
- Cell-type-specific marker identification
- Custom gene list support
- Dimensional reduction support (PCA, pre-computed embeddings)
- Cross-sample and cross-study benchmarking capabilities
- Customizable visualization tools (heatmaps, MDS, PCA plots)
- Hierarchical clustering analysis
- Comprehensive documentation with vignettes
- Examples for all 19 exported functions
- Support for optional dependencies (transformGamPoi, glmGamPoi)
- Efficient euclidean distance computation with custom C implementation
- Parallel processing support via BiocParallel
- Proper handling of sparse matrix formats
- Robust batch effect handling
- Consistent results across different label granularities
- Main vignette: comprehensive tutorial with real-world examples
- Quick start guide: minimal workflow for rapid evaluation
- Full API documentation for all exported functions