Skip to content

Latest commit

 

History

History
76 lines (55 loc) · 3.02 KB

NEWS.md

File metadata and controls

76 lines (55 loc) · 3.02 KB

scTypeEval News

Version 1.1.2 (2026-07-16)

New Features

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

Improvements

  • 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).

Bug Fixes

  • 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.

Version 0.99.30 (2026-04-07)

Improvements

  • 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.

Changes

  • Removed get.optimal_clustering from the public package workflow/documentation while it remains in development.

Version 0.99.21 (2024-01-27)

New Features

  • 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

Improvements

  • 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

Bug Fixes

  • Proper handling of sparse matrix formats
  • Robust batch effect handling
  • Consistent results across different label granularities

Documentation

  • Main vignette: comprehensive tutorial with real-world examples
  • Quick start guide: minimal workflow for rapid evaluation
  • Full API documentation for all exported functions