NEWS

---

postexportKinetics 0.99.0

Bioconductor submission development line. It is not a Bioconductor release. The released GitHub version remains 0.1.0 (tag v0.1.0, unchanged). The package has no DOI.

• Version 0.99.0, the pre-release version required for a first Bioconductor submission.
• New postexport_data_from_se(): converts an already state-resolved SummarizedExperiment into a postexport_data object, exactly as postexport_data() would for the equivalent wide table.
  • Layout A: events as rows, destructive samples as columns, time and replicate in colData, and four assays N, N_s, C, C_s.
  • It does not pair libraries, normalise, convert rMATS output or infer inclusion/skipping.
  • SummarizedExperiment is now in Imports.
• Vignette: "Introduction" and "Installation" sections, and a short SummarizedExperiment input section.
• The example-data generation script is distributed as inst/scripts/generate_postexport_example.R. It reproduces the shipped example data exactly.

postexportKinetics 0.1.0

First public release, published as a GitHub software release. It is not a Bioconductor release (Bioconductor is the intended future distribution channel), and the interface may still evolve. The package has no DOI.

Scientific reference

• The numerical core is ported verbatim from the frozen manuscript implementation.
  • Repository: bioinformatics-sannio/postexport-kinetics.
  • Tag manuscript-revision-v1.0, commit 65c3b7368fb7686bfde3dab857f98c393bb534c5.
  • Archive of the frozen manuscript implementation (not of this package): doi:10.5281/zenodo.22944109.
• Ported components:
  • four-state ODE model and transcription schedules;
  • trapezoidal interval balances;
  • covariance estimation, shrinkage, eigenvalue floor and propagation;
  • whitened NNLS fits of the full model (sigma_c >= 0) and the null model (sigma_c = 0);
  • the statistic T = max(0, RSS0 - RSS1) and the boundary rule;
  • Crank-Nicolson null trajectories;
  • the replicate-level generative bootstrap and the add-one p-value.
• Regression tests compare the package with fixtures generated from the frozen tag, on macOS and on Linux CI.

Inference

• postexport_data() and validate_postexport_data(): validated wide or long input. Missing states are errors, never zero-filled.
• postexport_control(): numerical and inferential settings with the frozen defaults (B = 1999).
• fit_postexport_model(): full and null fits with diagnostics.
• test_postexport_conversion(): bootstrap test of sigma_c = 0 against sigma_c > 0, with boundary information and RNG metadata.
• print() and summary() methods for all result classes.

Simulation and design diagnostics

• simulate_postexport_kinetics(): the manuscript simulator, with:
  • transcriptional onset and a common intervention time;
  • the NONE, SHUTOFF and PSEUDO_SHUTOFF regimes;
  • replicate variability and the benchmark assay-noise presets.
• check_operational_domain(): diagnostic comparison of a design with the manuscript's corrected factorial benchmark. It is not a calibration guarantee.

Batch analysis, multiple testing, ranking and plots

• Multi-event calls run sequentially with per-event seeds. An unexpected error in one event is recorded as status "event_error".
• adjust_postexport_pvalues(): explicit multiple-testing adjustment (Benjamini-Hochberg by default) over valid tests, with user-defined families. A single test is a family of size one.
• rank_postexport_candidates(): exploratory prioritisation with the manuscript's composite score. It is not inferential.
• as.data.frame() methods give tidy result tables with documented column roles.
• plot() methods (ggplot2) for fits, tests, result sets, simulations and operational-domain checks.

Documentation and data

• Vignette vignette("postexportKinetics").
• Synthetic example data postexport_example and postexport_example_truth, with a long-format copy in inst/extdata.

Compatibility and checks

• Declared minimum versions: R (>= 4.1.0), ggplot2 (>= 3.4.0), testthat (>= 3.1.7).
• Continuous integration: R 4.1 and oldrel-1 on Linux; the current R release on Linux, macOS and Windows; the Linux frozen-reference regression gate.

Not yet available

• Parallel execution.
• rMATS input conversion.
• Comparator methods.
• SummarizedExperiment input.