Changes in version 0.99.2                        

Changes made in response to the Bioconductor review (#186).

Significant user-visible changes

  - NADIA no longer sets the random seed; call set.seed() to reproduce
    results. The seed arguments have been removed, and
    pattern_profiler_analysis() replaces seeds with n_starts.
  - 61 helper functions used by the metric and benchmarking functions
    are no longer exported (reducing the number of exported functions
    from 111 to 50; 51 with the new preprocess_se()). The high-level
    functions continue to return all previously available outputs.
  - Protein score columns are PG.Score and PG.Score.RunWise_* in every
    format, instead of PG.Cscore; the original report metric is
    preserved in the nadia_labels attribute. Older tables and .nadia
    files can still be loaded.
  - imp_method = "missForest" now treats proteins as observations, as
    MsCoreUtils does; observations = "samples" retains the orientation
    used by NAguideR.
  - imp_method = "nbavg" has been removed, and bpca no longer accepts
    nPcs.

New features

  - process_proteomics() accepts a SummarizedExperiment through the new
    import function preprocess_se().
  - export_format also applies to the exported matrices.

Dependencies

  - limma, rrcovNA, pROC and MsCoreUtils move to Imports; highcharter,
    reactable and paletteer move to Suggests; readr is removed. The
    availability of each suggested package is checked before use, and an
    installation command is provided if it is missing.
  - Imputation methods zero, min, with, MinDet, MinProb, QRILC, bpca and
    missForest, along with vsn normalization, now use MsCoreUtils and
    produce the same results as before.

Bug fixes

  - Batch correction with protected covariates no longer discards
    measured values: proteins that cannot be corrected while protecting
    the covariates receive batch-only correction.
  - All plotting functions resolve palettes consistently; an unknown
    palette name raises an error.
  - Heatmaps accept partial row annotations, preserve condition colors
    in comparison-specific heatmaps and no longer modify ComplexHeatmap
    options.
  - Interactive tables on the same page no longer interfere with each
    other. Excel export preserves text that starts with digits, and
    condition names containing underscores are accepted.
  - Exported tables quote fields that contain tabs, newlines or quotes.
  - .nadia files restore the processing result exactly, except for the
    precision weights produced by limpa.
  - im_prepare_se() accepts norm_method = "log2".
  - A Parquet export without arrow fails before the analysis starts.

                       Changes in version 0.99.0                        

Initial Bioconductor submission.

  - Import of protein-level reports from Spectronaut
    (preprocess_spectronaut()), DIA-NN (preprocess_diann()) and Proteome
    Discoverer, including both TMT (preprocess_tmt()) and label-free
    (preprocess_lfq()).
  - process_proteomics() runs normalization (13 methods), optional batch
    correction, imputation (20 methods, including the two-stage MAR/MNAR
    combo strategy) and differential abundance analysis with limma or
    limpa.
  - Each protein's missingness is reported alongside its fold change:
    across the experiment, within the comparison and within each
    condition (MissGlobal, MissComp, MissCND1, MissCND2).
  - Evaluation of normalization and imputation methods
    (normalization_metrics(), imputation_metrics()) and spike-in
    benchmarking with OpDEA metrics (benchmarking_proteomics(),
    benchmarking_multiple()).
  - Soft clustering of abundance profiles with Mfuzz
    (pattern_profiler_analysis()).
  - Interactive and static figures (volcano plots, boxplots, PCA plots,
    heatmaps and cluster profiles) and interactive result tables.
  - write_nadia() and read_nadia() save and load a complete analysis in
    a single DuckDB file.
  - Example dataset nadia_dia and trimmed example reports in
    inst/extdata/.