NEWS

scTypeEval 1.1.2 (2026-07-16)

Added aggregation, center, and scale parameters to run_pca() function, allowing users to tune PCA computation for single-cell and/or pseudobulk data.

Added comprehensive test coverage for the aggregation parameter in run_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 selective aggregation parameter—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 = FALSE vignette chunks and aligned examples with the current API.
Updated reciprocal classification vignette examples to use reduction = FALSE, avoiding unsupported dimensional-reduction usage for recip_classif methods.
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_clustering from 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