DGEAR 0.2.1

Breaking changes

• read_and_preprocess_data() has been removed. Its responsibilities are now handled internally by each test function via the new .prepare_data() helper in utils.R. Existing scripts that called this function directly must be updated to pass dataframe, con1, con2, exp1, exp2 to the test functions or to DGEAR() instead.

New features

• Unified function interface: all five test functions (perform_t_test, perform_anova, perform_dunnett_test, perform_h_test, perform_wilcox_test) now share a consistent signature: (dataframe, con1, con2, exp1, exp2, alpha = 0.05, annot_df = NULL).

• Optional gene annotation (annot_df): users may supply a GEO SOFT-style annotation data.frame with columns ID and Gene.Symbol (or Gene.symbol) to map probe row names to gene symbols. When annot_df = NULL (the default), row names of dataframe are used directly — making the package fully usable without any annotation file.

• Richer DGEAR() return value: the main function now returns a named list with four elements:

  • DEGs — genes that passed majority voting
  • FDR_Table — per-test BH-adjusted p-values, Fisher-combined FDR, ensemble voting score, and log2FC for every gene
  • Results_Table — concise summary (G_Symbol, CombineFDR, log2FC, Ensemble)
  • IndividualTests — raw Table and DEGs output from each of the five test functions

• Fisher’s combined p-value: DGEAR() now computes a combined p-value across all five tests using metapod::parallelFisher and includes it as CombineFDR in FDR_Table and Results_Table.

• Internal helpers in utils.R: .prepare_data() centralises log2 normalisation and .resolve_symbols() centralises gene symbol resolution, eliminating code duplication across test files.

• Defensive Dunnett test: perform_dunnett_test() now silently assigns NA p-values to genes with zero variance, NA values, or infinite values, rather than throwing an error that would abort the full analysis.

• Input validation in DGEAR(): informative stop() messages are raised for invalid column ranges, alpha outside (0, 1), or votting_cutoff outside 1–5.

• 82-test testthat suite: tests/testthat/ now contains 7 test files covering structure, value sanity, biological correctness, defensive handling, alpha/cutoff sensitivity, and annotation behaviour for every exported function.

• Interactive demo script: inst/examples/DGEAR_demo.R is a fully annotated walkthrough callable via source(system.file("examples", "DGEAR_demo.R", package = "DGEAR")).

Bug fixes

• Eliminated the R CMD check NOTE “no visible binding for global variable ‘G_Symbol’” by removing all reliance on globally defined annotation vectors. Gene symbols are now always resolved locally inside each function.

• Removed library(DGEAR) from all \examples{} blocks in .Rd files; R CMD check loads the package automatically, so the call caused an error.

• Wrapped perform_dunnett_test() examples in \donttest{} to prevent R CMD check failures from the DescTools::DunnettTest external dependency in constrained check environments.

• fdr columns are now consistently typed as integer (0L / 1L) rather than a mix of NA and numeric 1.

DGEAR 0.1.4

Development version (0.1.3.9001) — changes after first release

• Improved FDR method across all statistical test functions.
• Improved function return types: each test now returns a list of two elements, the full statistical table and the significant genes.
• BH-adjusted FDR values for each test are now shown in a column named BH.

Changes in 0.1.4.9000

• perform_h_test(): removed the FC parameter; fold-change is now computed internally from the data.
• perform_wilcox_test(): internal improvements.

DGEAR 0.0.0.9000

NEWS.md setup

• Added NEWS.md with newsmd.