Redesigned interactive digest reports with ASCII-safe score profiles, explicit scoring settings and peptide counts, wrapped identifiers, and width-aware enzyme comparisons.
Preserved resolved scoring metadata through single-protein, comparison, batch, sensitivity, and plotting workflows.
Restored Windows protein-chunk parallelism through base R socket workers. Unix uses fork workers, and failed chunks on either backend are retried sequentially with a classed warning.
Validation and fixes
Standardized malformed-input handling across digestion, scoring, evaluation, export, reporting, diagnostics, and plotting. Invalid inputs now fail through package conditions before reaching base R errors.
Rejected duplicate protein identifiers and duplicate data-frame columns before they can collapse or corrupt downstream results.
Applied the zero-count digest rule consistently: digests without peptides inside the active length range receive a zero composite score and a Poor verdict.
Corrected the workflow-preset comparison to apply complete preset settings against an explicit three-protein digest background.
Distinguished theoretical products, length-valid peptides, combined comparison windows, model score bands, and model-based rankings throughout package-facing documentation.
Fixed missed-cleavage filtering in peptide-overlap plots when all levels are requested, and validated plot titles and figure dimensions before rendering.
Corrected weight sensitivity tie handling, corner diagnostics, batch aggregation, and single-result plot semantics. Results now record simulation settings, and batch memory is bounded by a smaller default chunk.
Testing
Reworked the test suite around public contracts, scientific invariants, classed conditions, metadata, plot semantics, and platform-specific batch execution.
Added generated-sequence properties and a malformed-input audit covering all 37 exported functions.
Added fast contract tests for shipped FASTA, tool-comparison, and PeptideAtlas artifacts. Data-generation workflows now run as separate offline audits and fail clearly when required inputs are missing.
Removed an architecture-dependent threshold collision from the diagnostics ablation oracle so installed-package tests retain the same verdict-flip expectation across BLAS implementations.
Corrected Windows-specific test assumptions about normalized path separators and avoided reinstalling an already installed package during R CMD check.
Documentation and metadata
Prepared the package metadata and installation guidance for initial Bioconductor submission.
Reconciled the README, vignettes, help pages, package metadata, and generated site with the current scoring and validation contracts.
Removed the redundant custom package citation because pepVet does not yet have an associated publication DOI. Installed citations now come from DESCRIPTION metadata.
Described the scoring weights as documented expert priors because their original derivation record is not retained. The numeric weights are unchanged.
Release infrastructure
Added Linux coverage reporting with a 90 percent floor, retained coverage artifacts, and Codecov upload support.
Added pkgdown builds for pull requests and GitHub Pages deployment from main, with bounded build stages and cancellation of obsolete runs.
Added a source-package artifact build alongside parallel release-R checks on Linux, macOS, and Windows, with unfinished checks cancelled after the first check failure.
Split CI installation into role-specific pak dependency plans and preserved resolved-library caches when later workflow stages fail.
Bounded dependency installation in every workflow so a stalled package resolver fails promptly instead of consuming an entire job timeout.
Consolidated routine Linux build, check, coverage, and pkgdown work behind one uncached dependency installation while retaining parallel macOS and Windows checks. This avoids intermittent cache-service stalls across duplicated Linux setup jobs.
Added version-tag release-candidate builds and tag/published-release BiocCheck workflows. Release artifacts are built in temporary storage and are not published automatically.
pepVet 0.1.7
Empirical concordance
Added a reproducible PeptideAtlas concordance analysis. In the sampled human tryptic digests, peptides inside the default length and GRAVY windows had a 38.0 percentage-point higher mean observation rate than peptides outside them. The result is limited to the sampled data and is not an experimental calibration of pepVet scores.
Added bounded grid-search and threshold-calibration analyses, documented their limits, and retained the current defaults as strict, conservative settings by design.
Corrected the concordance artifacts to store full protein lengths and regenerated the committed CSV and RDS files.
Package quality
Standardized docstrings, limitations, error handling, plot conventions, package-qualified calls, comments, and line formatting across the source.
Unified GRAVY calculation under one vectorized helper and removed stale helper references from package and data-generation code.
Removed inactive GitHub workflow files and generated vignette data from version control so the repository matches the current release process.
Fixes
Fixed edge cases in row binding, cleavage-efficiency summaries, GRAVY missing-value handling, and sensitivity plot validation.
Fixed recommend_enzyme() tie handling so all enzymes within tolerance are returned in alphabetical order.
Corrected charge scoring for single-residue peptides so terminal basic residues are not counted as internal charge sites.
Fixed plot configuration updates so invalid changes leave the active configuration unchanged.
pepVet 0.1.6
New functions
score_diagnostics() runs VIF, PCA, and ablation analyses on a batch_evaluate() result to quantify multicollinearity, dimensionality, and component importance in the scoring model.
plot_score_diagnostics() visualises the diagnostics result as a three-panel figure: VIF bar chart with severity-colored bars, PCA scree plot with cumulative variance, and ablation waterfall with error bars and verdict-flip counts.
New documentation
Added Score Diagnostics for pepVet Scoring Models article covering VIF, PCA, and ablation analysis with worked examples.
Added pepVet in the Tool Landscape article comparing pepVet against MS-Digest, ExPASy PeptideMass, Protein Cleaver, ProteaseGuru, and PeptideRanger.
Added diagnostics cross-reference to the introduction vignette.
Bug fixes
Fixed sensitivity_analysis() batch-mode verdict instability: now uses full 3-level classification (Good/Moderate/Poor) instead of checking only the Good boundary.
Fixed vignette pipe example where recommend_enzyme() received a comparison tibble instead of a protein sequence.
Fixed tool-comparison vignette: replaced fragile ../inst/extdata/ paths with system.file().
Fixed verdict threshold table in introduction vignette: said 0.7 / 0.4, code uses 0.65 / 0.40.
Housekeeping
Refactored zzz.R to avoid the lockBinding R CMD check NOTE while preserving active bindings for mutable plot configuration.
Fixed README installation instructions (removed S4Vectors from BiocManager install; pulled in automatically by Biostrings).
Enabled automatic CI triggers for R CMD check on push and PR across ubuntu, macOS, and Windows.
Added .gitattributes, .Rbuildignore updates, .onAttach removal, and cli cleanup.
Docstring standardisation and .bind_rows() type-safety improvements.
pepVet 0.1.5
New functions
sensitivity_analysis() performs Monte Carlo weight perturbation via Dirichlet sampling. Reports simulated verdict frequencies, composite intervals, rank stability (top-1, Kendall tau), per-protein instability in batch mode, weight importance (R squared), and corner-case composites. Optional importance and corner_cases diagnostics.
plot_weight_sensitivity() generates a verdict-coloured density ridge plot from a sensitivity analysis result, with threshold lines and a rug mark at the default composite score.
Scoring
Zero-cleavage hard-fail: proteins with no cleavage sites for a given enzyme now receive composite = 0 and verdict = “Poor” regardless of other scores. Previously an uncuttable protein with ideal length/hydro/charge could score up to 0.426, crossing the Moderate threshold. The fix belongs in the scoring engine so sensitivity analysis treats these as deterministic negative controls.
Plot fixes
plot_digest_profile() coverage panel now filters to the exact requested MC level instead of showing all levels. Overlapping peptides are stacked into sub-tracks via greedy interval packing, and the panel title includes the MC level. Gap overlays remain but the subtitle no longer counts uncovered regions; a small italic caption notes when some gaps are too narrow to render at plot scale.
Site
Updated pkgdown color palette to match the new pepVet logo colors. Primary/secondary swapped for better link contrast. Favicons regenerated.
Replaced all logo references with the new pepVet-logo.png. Navbar shadow, rounded code blocks, sticky TOC sidebar added.
Flattened the Articles dropdown. Removed redundant “Getting Started” entry and section headers.
Added prev/next navigation buttons to reference pages (client-side JS).
pepVet 0.1.4
Housekeeping
Standardised cli error messages across the package. Replaced all paste0("{.arg ", var, "}") patterns with "{.arg {var}}" cli glue syntax in .abort(), cli_warn(), and cli_inform() calls. Also converted "i" bullet paste() calls to {.val {x}} formatters.
Added .bind_rows() helper and replaced all bare do.call(rbind, ...) calls with it. The helper returns an empty tibble for empty input instead of failing silently.
Replaced import(cleaver) with @importFrom cleaver cleavageRanges. Removed the .cleavage_ranges workaround via get("cleavageRanges", envir = asNamespace("cleaver")) and now calls cleaver::cleavageRanges() directly.
Removed dead code: .compute_difficulty_flags() in evaluation.R was defined but never called. Its logic was superseded by .batch_difficulty_flags().
Made plot_coverage_map() gradient stops (color_by = "hydrophobicity") data-driven instead of hardcoding the GRAVY valid-range boundary at 0.6. The 4 color stops are now evenly spaced across the actual GRAVY range of the displayed peptides.
Suppressed three R CMD check notes by adding .lintr, tmp/, and paper/ to .Rbuildignore.
Added missing @return roxygen tags to 10 internal helper functions across the source.
Removed duplicate .classify_verdict() from plot_utils.R. The function was defined in both scoring.R and plot_utils.R with identical logic; the duplicate silently shadowed the original due to file-sourcing order.
Fixed plot_batch_comparison() heatmap colorbar to use explicit unit(..., "pt") instead of numeric (lines) values.
Bug fixes
Fixed .onLoad() failure when called on an already-loaded namespace. The rm() call on locked namespace bindings was guarded with bindingIsActive() checks so the active-binding installation runs exactly once.
pepVet 0.1.3
Defaults
Changed default missed_cleavages from 0L to 1L across all evaluation functions. MC=1 reflects standard bottom-up proteomics practice; MC=0 produces unrealistically poor scores for every enzyme.
Changed default scoring weights to documented expert-prior values: S_length = 0.200, S_coverage = 0.348, S_count = 0.226, S_hydro = 0.138, S_charge = 0.088. The derivation record needed to support the earlier method and consistency-ratio description was not retained, so those claims have been removed.
Lowered the Good verdict threshold from 0.70 to 0.65 to better align with the observed score distribution under the new weights. Centralised verdict thresholds in .pepvet_params accessed via .get_param().
Performance
digest_protein() rewritten for batch workloads. .cleavage_ranges() is now called once on the full AAStringSet instead of once per protein, eliminating repeated S4 dispatch overhead. The non-efficiency path pre-allocates six output vectors and fills them in a single loop before constructing one tibble, replacing ~20 K individual tibble builds followed by do.call(rbind, ...).
batch_evaluate() restructured around two bulk calls, one digest_protein() and one score_peptides() for the entire input, instead of a per-protein loop over evaluate_digest(). A new internal helper .batch_difficulty_flags() computes all four difficulty flags across all proteins simultaneously via tabulate(), tapply(), and a vectorized GRAVY calculation, replacing ~20 K per-protein subset/GRAVY/tibble::as_tibble() cycles.
triage_proteins() fully vectorized. Row-by-row vapply logic replaced with nested ifelse() and a rowSums(matrix < 0.5) component check, making the function O(n) in a single pass.
S_coverage scoring no longer uses IRanges::reduce(). Overlapping intervals are now merged inline via order() + cummax() on sorted starts and ends. Same result with no S4 dispatch.
Scoring helpers (S_coverage, S_count, S_hydro, S_charge) accept a pre-computed valid_digest argument so callers can extract valid peptides once and share it across all four components instead of calling .extract_valid_digest() four times per protein.
New functions
Added batch_compare_enzymes() to score an entire proteome against multiple enzymes in one call, returning a tidy tibble of class pepvet_batch_comparison with one row per protein-enzyme pair. Parallel execution is supported via parallel::mclapply on Unix (fork copy-on-write, zero serialization overhead) and parallel::parLapply on Windows; both are part of base R. Proteins are split into equal chunks, so all cores workers are utilised regardless of the number of enzymes.
Parallel robustness
Fixed a bug in batch_evaluate() where parallel::mclapply worker crashes produced silently corrupted results. Failed chunks are now detected and retried sequentially.
Amino acid data
Added U (selenocysteine, Sec) to the aa_properties reference table. 25 human proteins contain selenocysteine and were previously rejected during input validation.
Added O (pyrrolysine, Pyl) to the aa_properties reference table (row 22). Monoisotopic mass verified from PubChem CID 119813 and ChEBI CHEBI:91273 (C₁₂H₂₁N₃O₃, 255.15829 Da). Hydrophobicity and pKa are NA: the Kyte-Doolittle scale predates the discovery of pyrrolysine (1982 vs 2002), and the ε-amino group is tied up in an amide bond and is not titratable. calculate_gravy() now passes na.rm = TRUE so sequences containing O return a valid GRAVY score computed from the remaining residues.
Removals
Removed plot_enzyme_protein_heatmap(). The 2D tile matrix did not add enough over the other comparison functions (plot_enzyme_comparison(), plot_batch_comparison()) to justify maintenance.
Removed plot_protein_comparison(). The grouped bar chart did not add enough insight beyond plot_enzyme_comparison() and the batch-level comparison functions.
pepVet 0.1.2
Visualization
Added plot_digest_profile() as a four-panel single-protein diagnostic showing length distribution, GRAVY scatter, coverage map, and component scores.
Added plot_coverage_map() for horizontal sequence coverage visualisation with valid/invalid peptide segments and missed-cleavage expansion lanes.
Added plot_batch_summary() for two-panel proteome overview: verdict histogram and score-versus-length scatter.
Added plot_proteome_heatmap() for hierarchical clustering of component scores across a batch. Requires pheatmap (guarded).
Added plot_component_scatter() for 2D scatter of any two score components across all batch proteins.
Added shared infrastructure in R/plot_utils.R: .pepvet_pal colour palette, .pepvet_theme() base theme, and reusable panel builders.
Housekeeping
Established coding conventions and standardisation strategy for consistency across the package.
pepVet 0.1.1
Batch evaluation, triage, and export
batch_evaluate() now returns a flat tibble with one row per protein instead of a named list of evaluate_digest() results. Columns: protein_id, protein_length, n_peptides, n_valid_peptides, all component scores, composite_score, verdict, median_peptide_length, and four sequence-level difficulty flags. This is a breaking change for any code that accessed results via batch[[protein_id]]$scores.
Added summarize_batch() to extract aggregate statistics from the batch tibble: verdict distribution, composite score distribution, per-component means, bottom-10% problem proteins, and heuristic enzyme-switch candidates.
Added triage_proteins() to append an action column ("proceed", "consider_alternative", "try_other_enzyme", "skip") to the batch tibble based on verdict and sequence-level difficulty flags.
Added pepvet_check() as a single-call convenience wrapper: evaluates a protein digest and immediately prints a styled console report. Returns the evaluate_digest() result invisibly.
Added export_peptide_list() in a new R/export.R module. Exports valid peptides in "skyline" (precursor-charge CSV with computed m/z for Skyline import), "generic" (annotated CSV with GRAVY, pI, and validity), or "fasta" format. Writes to a file when file is specified; returns a tibble or character vector otherwise.
Fixed export_peptide_list() generic format to include a pI column alongside gravy and valid.
pepVet 0.1.0
Digestion and annotations
Added annotate_cleavage_sites() for sequence-local cleavage-efficiency annotation of trypsin-family digestion sites.
Added optional cleavage_efficiency output in digest_protein() so peptide tables can carry high/medium/low cleavage-risk context without changing the default schema.
Added n_high_efficiency_sites and n_low_efficiency_sites to evaluate_digest() output as informational protein-level summaries.
Scoring and utilities
Added peptide mass and pI utilities plus residue-level monoisotopic mass data for fractionation-aware planning workflows.
Retained preset tracking through preset_used so score tables can distinguish shipped workflows from custom configurations.
Documentation
Reworked the README and core articles for the 0.1.0 release so digestion, scoring, presets, and cleavage annotations are explained as one coherent planning workflow.
pepVet 0.0.4
Scoring
Reworked S_count to use an enzyme-aware expected peptide length instead of a fixed trypsin-derived constant.
Added median_peptide_length to scoring output so enzyme comparisons show the denominator behind S_count.
Added configurable gravy_range and length_range arguments across the scoring and evaluation helpers.
Added pepvet_preset() with presets for standard DDA, DIA, targeted assays, membrane proteomics, FFPE-style degraded samples, and fractionated workflows.
Clarified that S_charge tracks extra internal basic-residue richness, not baseline peptide ionizability.
Added preset_used metadata to scoring output and evaluate_digest() params so named presets and custom scoring configurations can be distinguished explicitly.
Documentation
Rewrote the package website copy, docstrings, README, and articles for v0.0.4 with detailed workflow guidance and preset-specific examples.
Added a dedicated article on workflow presets, including practical examples and use-case notes for each preset.
Added scoring-model positioning, evidence-basis, scope, and known-limitations guidance so heuristic and literature-backed assumptions are documented in-package.
Site
Updated pkgdown navigation to surface the preset guide alongside the core workflow articles.
digest_report() prints styled console output for evaluation and comparison results, with colour-coded bar charts and ranked tables.
Documentation
Rewrote README.md with scoring component table, reference fixture table, and workflow diagram.
Expanded the getting-started vignette to cover all seven exported functions end-to-end.
Added Choosing a Proteolytic Enzyme article covering enzyme biology, worked comparisons on BSA, Histone H3.1, and alpha-synuclein isoforms, and guidance for membrane proteins, phosphoproteomics, and IDPs.
Added Understanding the Scoring Model article with mathematical definitions for all components, weight customisation guidance, verdict calibration notes, and known limitations.
Site
Configured pkgdown site with structured navbar, grouped reference index, and flatly Bootstrap 5 theme.
pepVet 0.0.2
Core engine
digest_protein() performs cleaver-compatible digestion with validated input handling for character sequences, AAString, AAStringSet, and FASTA paths.
score_peptides() computes five component scores (S_length, S_coverage, S_count, S_hydro, S_charge) and optional proteome-aware S_unique.
Data
Added aa_properties reference tibble with Kyte-Doolittle hydrophobicity, molecular weight, and side-chain pKa values.
Added eight reference FASTA fixtures in inst/extdata/.
CI
Established GitHub Actions workflows for R CMD check and lint.
