NEWS

---

darwin 1.0.0

Initial Release

Features

• Multi-objective gene selection using NSGA-II algorithm
• Built-in objective functions:
  • compute_correlation(): Minimize pairwise correlation between cell types
  • compute_distance(): Maximize pairwise Euclidean distance
  • compute_condition(): Minimize condition number for stability
• Support for custom objective functions
• Two selection modes:
  • standard: Variable number of genes
  • fixed: Fixed number of genes
• Flexible input formats:
  • Seurat objects (V4 and V5)
  • SingleCellExperiment objects
  • Matrix and data.frame
• Multiple selection methods from Pareto front:
  • Weighted selection
  • Index-based selection
  • Target value proximity
• Built-in deconvolution methods:
  • NuSVR (requires e1071 package)
  • NNLS (non-negative least squares)
  • Linear regression
• High-performance implementation:
  • Vectorized R operations
  • Optional C++ acceleration via Rcpp/RcppArmadillo
  • Parallel computing support via future package
• Cross-platform compatibility (Windows, macOS, Linux)
• Comprehensive test suite
• Detailed documentation and vignettes

Technical Details

• R6 class-based interface for object-oriented usage
• Caching of fitness evaluations for efficiency
• Configurable genetic algorithm parameters
• Progress tracking with cli package

---

This package implements the multi-objective gene selection methodology for RNA-seq deconvolution.