Changes in version 0.99.3                        

Bugfixes and package/documentation updates for Bioconductor submission.

Bug fixes

  - ani_distance() and plot_mds() failed when a taxonomy was supplied:
    the function body referred to an undefined object taxa, which is now
    fixed.
  - define_priors(add_dispersion_prior = TRUE) errored because the
    internal helper was called without a required argument. Now the
    documented normal(0,1) dispersion prior is hardcoded.
  - plot_manhattan() passed its default method = "holm" to hadjust(),
    which only implements HMP and Bonferroni. Previously, this silently
    fell back to Bonferroni. The default is now NULL, which resolves to
    "HMP" when a taxonomy is supplied and "holm" otherwise, and
    hadjust() warns when an unsupported method is given. Default
    taxonomy Manhattan plots now use HMP.
  - plot_volcano(contig_names = ). This input was never used and the
    supplied names now label the significant hits.
  - Updated all fitters to record scale_continuous status in the
    returned strainspy_fit object.
  - geom_segment() layers updated to use linewidth instead of size,
    avoid ggplot2 deprecation warning.

API

  - Added new accessors for strainspy_fit: getRowData(),
    getCoefficients(), getStdErrors(), getPValues(), getZIStdErrors(),
    getZIPValues(), getVcov(), getZIVcov(), getDesign(), getFamily(),
    getFitCall(), getConvergence() and isScaleContinuous().
    strainspy_priors: getPriorsDf(). Avoiding the need to use @ and
    slot() calls on S4 objects.
  - Fixed typos: est_disperion_prior is now est_dispersion_prior in
    compute_eb_priors(), and scale_continous is now scale_continuous in
    glmQBFit().
  - hadjust() always required taxonomy and the NULL default was removed.
  - Unused internal functions filter_by_presence_and_rescale(),
    filter_by_presence_longitudinal() and getDispEst() were removed.
  - CInternal helpers no longer generate help pages.

Documentation

  - The vignette now has an Installation section with BioConductor
    instructions, and the README a containere section.
  - Many minor help-page corrections: argument defaults, swapped
    descriptions, stale function names and typos.

                       Changes in version 0.99.2                        

Minor bug fixes and documentation updates.

  - Added installation instructions to README for R-universe and
    Bioconda
  - Added tests to improve coverage of simulation_helpers.R and others.
  - Removed redundant getFDR() function from strainspy_fit-class.R and
    documentation.

                       Changes in version 0.99.1                        

Bug fixes

  - R.utils moved from Suggests to Imports. data.table::fread() stops
    with an error when asked to read a gz or bz2 file and R.utils is not
    installed, so read_sylph(), read_sourmash(), read_metaphlan() and
    read_taxonomy() all failed on compressed input for anyone who
    installed without the suggested packages. Every example file shipped
    in inst/extdata is gz compressed, so this affected the whole
    documented workflow. A minimum version of 2.13.0 is required because
    fread() refuses older versions of R.utils on R 4.5 and above.
  - 1:n replaced with seq_len() and seq_along() throughout. On an empty
    input 1:0 counts backwards and yields two indices rather than none,
    so zero-row data produced an error or a spurious extra iteration
    instead of an empty result.

Dependencies

  - The dependency on R has been relaxed from 4.6.0 to 4.1.0. The native
    pipe is the only version-dependent feature the package uses. The
    previous constraint prevented installation on the R release before
    current, which R-universe builds and checks, and made the package
    impossible to install alongside the Bioconductor packages
    distributed by bioconda, which are built against an earlier version
    of R.

Other changes

  - Added inst/CITATION, so citation("strainspy") now returns the
    StrainSpy preprint (doi:10.64898/2026.08.30.748153).
  - The redundant print() method for strainspy_priors has been removed.
    A show() method was already defined and is more informative, and
    print() on an S4 object dispatches to it, so print(priors) now
    reports more than it did before.
  - Error conditions no longer wrap their message in paste(), and the
    names of missing contigs are now included in the condition itself
    rather than printed separately to standard output, where they could
    become detached from the error.
  - Continuous integration now runs the R-universe workflow that
    Bioconductor uses to build and check submissions, replacing a
    narrower R CMD check matrix and a separate BiocCheck job.
  - The README documents installation from R-universe as an alternative
    to installing from GitHub.

                       Changes in version 0.99.0                        

Version set to 0.99.0 in preparation for Bioconductor submission;
Bioconductor requires a 0.99.z version for new packages and increments
it to 1.0.0 at the next release.

Bug fixes

  - hadjust() no longer fails when taxonomy is given as a file path. The
    resolved taxonomy table is now passed to the internal ordering step,
    rather than the raw argument.
  - glmZiBFit() and glmObFit() now honour MAP_prior = NULL, the
    documented unpenalised maximum-likelihood fit, instead of failing on
    a NULL prior.
  - resolve_priors() now signals an error for an unrecognised MAP_prior.
    Previously any unrecognised value silently fell through to an
    unpenalised fit, so a typo, factor or list quietly changed the model
    that was fitted.
  - Failed per-feature fits are now identified by a single shared
    failed_fit_index() helper across all four fitting functions, which
    handles both backend failure conventions. The previous per-function
    checks each covered only one convention.

New validity checks

  - strainspy_fit gained a validity method requiring every per-feature
    slot to be aligned with row_data, so a length mismatch can no longer
    silently attribute one strain's statistics to another.
  - strainspy_priors gained a validity method checking method and the
    columns glmmTMB requires in priors_df. A malformed prior previously
    produced a fit object containing zero features with no error
    surfaced.
  - read_sylph(), read_metaphlan() and read_sourmash() now stop when
    sample or feature names are duplicated, which could previously
    happen when clean_names = TRUE reduced two input paths to the same
    basename.

Dependencies

  - readr is no longer required; metadata is read and written with
    utils.
  - gamlss and gamlss.dist moved to Suggests, as they are only needed
    for the non-default glmZiBFit(method = "gamlss") backend.
  - graphics, grDevices, tools and utils are now declared in Imports,
    and functions previously registered with globalVariables() are
    declared with @importFrom.

Other changes

  - The unused assay slot has been removed from strainspy_fit. It was
    never populated, so fit@assay always returned an empty matrix.
    Per-feature metadata is available from row_data.
  - Progress output is now emitted with message() rather than cat(), so
    it can be silenced with suppressMessages().
  - Example data in inst/extdata has been reduced to keep the source
    package within the Bioconductor size limit. The sylph profile and
    GTDB taxonomy are unchanged; the sourmash example is now a 20-sample
    subset.

                        Changes in version 0.1.0                        

  - Initial public release.
  - Added expanded test coverage across readers, modeling workflows,
    outputs, plotting, and utilities.
  - Added Bioconductor readiness updates: metadata enhancements,
    vignette scaffolding, and documentation improvements.