Changelog

perturbplan 0.3.1

Documentation

• Fixed outdated roxygen docs for pilot-data validators (validate_custom_baseline_rds, validate_custom_library_rds, validate_combined_pilot_data) to reflect current baseline_expression_stats + library_parameters schema
• Corrected extract_fc_expression_info and extract_expression_info docs: gene_list uses importance sampling, not “all genes”
• Updated get_pilot_data_from_package docs with all 8 biological systems and mapping_efficiency return value
• Fixed reference_data_processing docs: n_threads = NULL auto-detects threads, h5_only = TRUE returns NULL for baseline_expression_stats
• Fixed find_optimal_cost_design docs: cost_grid_size default is 50, not 200

Bug Fixes

• Fixed reference_data_preprocessing_10x: h5_rough = TRUE path now correctly respects skip_mapping_efficiency = TRUE

---

perturbplan 0.3.0

Breaking Changes

• New library estimation backend: Replaced minpack.lm NLS fitting with preseqR ZTNB/RFA saturation curve methods. library_estimation() now returns an rSAC_fn_wrapper list instead of scalar parameters.
• Dependency changes: Added preseqR and PoissonBinomial; removed minpack.lm
• Internal C++ functions now accept rSAC_fn_wrapper instead of UMI_per_cell/variation
• Removed deprecated downsampling_ratio and D2_rough arguments

New Features

• power_range parameter for cost_power_computation()
• Lower/upper caps for cells and reads in grid search
• Saturation threshold increased to 98% for better extrapolation
• Feature type filtering in obtain_qc_read_umi_table() (Gene Expression only)

Data Updates

• All 8 reference datasets regenerated with new preseqR-based library estimation

Documentation

• Expanded preprocess-reference vignette with multi-dataset saturation curve comparison
• Fixed LaTeX errors in PDF manual generation
• Rebuilt pkgdown website with updated favicon

Testing

• Test coverage increased to ~70% with 6 new test files
• Over 1,900 new lines of test code

Bug Fixes

• Fixed pilot data preprocessing bugs
• Fixed path handling bugs
• Improved numerical stability in dispersion estimation

---

perturbplan 0.2.0

Major Changes

Parameter Naming Standardization

• BREAKING CHANGE: Renamed reads_per_cell and raw_reads_per_cell parameters to sequenced_reads_per_cell in some functions
  • This change clarifies that the parameter refers to raw sequencing reads (before mapping), not mapped reads
  • Affects: compute_power_plan(), compute_power_posthoc(), cost_power_computation(), find_optimal_cost_design(), and related functions
• Improved internal handling of mapping efficiency with automatic conversion and rounding

New Vignettes and Documentation

• New vignette: “Prepare Data For Web App” (preprocess-reference.Rmd)
  • Comprehensive guide for preprocessing Cell Ranger outputs
  • Documentation for reference_data_preprocessing_10x() and reference_data_processing()
  • Examples for both Perturb-seq and TAP-seq workflows
  • Mathematical details of negative binomial expression model and read-UMI saturation curve
• Enhanced prospective-power vignette with updated parameter names
• Improved documentation for all pilot data preprocessing functions

Enhanced Pilot Data Preprocessing

• Added gene_list parameter to reference_data_processing() for TAP-seq targeted gene panel support
• Created example TAP-seq pilot data with targeted gene list
• Improved handling of different prior values for variation parameter (D2_rough)
• Enhanced example raw data to reduce numerical singularity issues

Bug Fixes

• Fixed Matrix package warnings for sparse matrix operations
• Resolved devtools::check() issues
• Fixed parameter passing in cost optimization functions
• Improved numerical stability in dispersion estimation

Website and Documentation Improvements

• Redesigned pkgdown site structure and homepage
• Added hex logo
• Rebuilt all documentation with updated parameter names
• Improved function examples and cross-references
• Enhanced vignette figures and visualizations

Testing

• All tests updated for new parameter naming convention
• Enhanced test coverage for parameter validation
• All R CMD check tests passing (0 errors, 0 warnings)

---

perturbplan 0.1.0

New Features

Cost Optimization Functions

• cost_power_computation(): Comprehensive power analysis with cost minimization across multiple experimental parameters
• find_optimal_cost_design(): Binary search optimization to find experimental designs meeting power targets
• obtain_fixed_variable_constraining_cost(): Helper function for cost-constrained experimental design

Enhanced Parameter Support

• Extended minimizing_variable parameter validation to support:
  • "TPM_threshold" - Expression threshold optimization
  • "minimum_fold_change" - Effect size optimization
  • "cells_per_target" - Cell count optimization (cost_power_computation only)
  • "reads_per_cell" - Sequencing depth optimization (cost_power_computation only)
  • "cost" - Total cost optimization
• Differentiated validation between cost_power_computation and find_optimal_cost_design

Documentation Improvements

• Added comprehensive @examples sections to all major functions
• Enhanced function documentation with detailed parameter descriptions
• Added @keywords internal annotations for helper functions

Bug Fixes

Critical Fixes

• Fixed minimum_fold_change < 0.6 bug: Added safety check and fallback mechanism when cell range identification fails due to small effect sizes
• Parameter naming standardization: Changed all tmp_threshold variables to TPM_threshold throughout codebase to prevent typos

Validation Improvements

• Enhanced input validation for cost optimization functions
• Added NSE-safe dplyr operations using .data$ notation
• Improved error messages and parameter checking

Performance Improvements

• Optimized switch statement logic for new minimizing variables
• Efficient handling of cost calculations across experimental parameter grids
• Enhanced C++ integration for computationally intensive operations

Package Structure

• Updated .Rbuildignore and .gitignore for pkgdown support
• Comprehensive test coverage for all new cost optimization features
• Maintained backward compatibility with existing power analysis functions

---

Breaking Changes: None - all new features are additive and maintain full backward compatibility.

Dependencies: No new dependencies added for core functionality.