NEWS

---

RankMap 0.99.1

Changes

• Updated DESCRIPTION to require R (>= 4.6.0).
• Refactored repeated matrix transposition logic in trainRankModel() and predictRankModel() into an internal helper function (.transposeMatrix).
• Reduced use of :: by importing selected functions (e.g., glmnet(), cv.glmnet()) to improve dependency checking.
• Added package-level documentation (RankMap-package.R).
• Updated vignette:
  • Removed GitHub installation instructions.
  • Added Bioconductor installation instructions using BiocManager::install("RankMap").
  • Added unique labels to all code chunks.

RankMap 0.99.0

Initial submission to Bioconductor

New Features

• Fast, robust, and scalable reference-based cell type annotation using multinomial regression on ranked expression matrices.
• Supports both single-cell and spatial transcriptomics data.
• Compatible with Seurat, SingleCellExperiment, and SpatialExperiment objects.
• Core function RankMap() provides a streamlined pipeline for preprocessing, model training, and prediction.
• Customizable preprocessing: top-K gene masking, optional binning, expression weighting, and scaling.
• Additional functions:
  • computeRankedMatrix() – generate ranked matrices
  • trainRankModel() – train multinomial GLM
  • predictRankModel() – apply trained model to query data
  • evaluatePredictionPerformance() – assess accuracy
• Optimized for large datasets with significantly faster runtime than SingleR, Azimuth, and RCTD.