Changelog

Source: NEWS.md

scop (development version)

• fix: RunRCTD(), RunCARD() and RunSPOTlight() align reference labels and metadata to selected-assay cell IDs before filtering, including assays containing only part of a Seurat object. Ambiguous IDs are rejected. RCTD preserves full selected-assay library depths through analysis-gene selection; zero-count filtering still uses the selected shared features.

• fix: GetSpatialResult() returns current-object cells in their current order and recomputes BANKSY and PRECAST count summaries after subsetting. New BANKSY fits and detailed SmoothClust, CHOIR, RareQ and PRECAST results retain private per-cell identities, including same-ID-set renames. SmoothClust cell-column storage and CHOIR explicit cell IDs are handled correctly. Missing or invalid provenance reports a clear error. Stored native fit results remain unchanged, and compatible legacy metadata reads are retained.

• refactor: Remove unused internal helpers and Rcpp bridges, share spatial neighborhood filtering and heatmap border preprocessing, and drop redundant Imports checks and obsolete ggplot2 compatibility code.

• change: Migration: RunBANKSY(assay = NULL) now prefers the selected image’s assay, then Spatial, RNA, and DefaultAssay; set assay explicitly to reproduce analyses that used a different assay.

• change: Migration: RunBANKSY() now reports its actual inputs and result locations and GetSpatialResult() reads clusters, parameters, and summaries; use verbose = FALSE to suppress receipts or store_results = FALSE to omit detailed stored results. A lightweight retrieval index still records custom cluster columns and exact sample membership, including overlapping names such as S1 and S1_A. Sample-specific access to legacy metadata without an index requires rerunning BANKSY; cluster labels alone cannot identify samples safely.

• change: Migration: set sample.by to fit BANKSY separately by sample; combined domain labels are sample-prefixed and are not aligned across samples.

• fix: C++ gene-set scoring uses GSVA >= 2.6 defaults when the optional GSVA package is absent, avoiding version-probe errors in sparse RunGSVA() z-score and PLAGE scoring. Installed GSVA versions retain their existing standardization and sparse-walk behavior.

• fix: RunDEtest() and native spatial APIs now distinguish an omitted object from an explicitly supplied object = NULL when resolving the deprecated srt= alias. A non-NULL srt= supplied alongside object follows the shared duplicate-input error behavior. The alias remains supported with a deprecation warning until scop 1.0.0, when its removal is scheduled.

• fix: RunGSVA(), CellScoring(), RunDynamicEnrichment(), RunCellRankEnrichment(), AnnotateFeatures() and RunMetabolism() normalize Latin species names before database lookup. Lowercase names and space, dot or hyphen separators now resolve to the same annotations as Homo_sapiens and Mus_musculus, avoiding missing-input errors and silently skipped databases.

• chore: Remove Rfast from Suggests after confirming it has no SCOP source, test, documentation, native, build-script, or CI use.

• change: RunPropeller() now uses speckle’s transformed proportions and empirical Bayes moderated tests instead of the internal logit t-test. It supports transform, robust, trend, records backend results and designs, and blocks on donor IDs for fully paired comparisons. At least two samples per condition are required. SCOP bootstrap intervals remain descriptive intervals for the untransformed proportion ratio.

• fix: RunCytoSPACE() samples singleton reference types without selecting cells from other types, preserves cell labels that collide after metadata name conversion or with the total-count column, and reports no dominant type for zero-cell spots. Empty assignments are supported; stored results record the collision-safe metadata suffixes and total-count column name.

• change: Database preparation now exposes independent PrepareGO(), PrepareKEGG(), PrepareMSigDB() and other source interfaces. PrepareDB() dispatches to these interfaces, with shared annotation caching and conversion. Combined model and custom annotation requests retain their existing structure. Chromosome mappings are resolved from the annotation namespace on first use.

• feat: Add experimental IREA analysis and plots with unified object inputs, explicit contrast provenance and checksum-checked reference preparation through PrepareDB(db = "IREA_<cell type>"), alongside annotation databases. Plots reuse plot_type and padjustCutoff; diagnostics use the shared logging interface. Numerical equivalence to the official portal remains unestablished.

• change: label_insitu now defaults to TRUE in CellDimPlot(), FeatureDimPlot(), PAGAPlot(), and CCC embedding-network plots, so label = TRUE shows group names (or cluster IDs) directly on the plot instead of renumbered labels that must be matched through the legend. Set label_insitu = FALSE to restore the numbered labels.

• fix: Differential-abundance plots now project Milo neighborhoods through stored member cells, use SpatialFDR for Milo neighborhood significance, and use scCODA credibility for scCODA plots. Milo group summaries and the internal Propeller-style test now report their statistical provenance.

• fix: Sample-level abundance tests preserve sample-condition units and support fully paired donor IDs in Milo and the internal Propeller-style test. Permutation testing now honors seed.

• change: Milo, scCODA, and Propeller require biological sample.by by default. allow_pseudo_samples = TRUE retains virtual samples for descriptive effects, with inferential statistics marked unavailable.

• fix: ProportionTestPlot() accepts result_level = "neighborhood" for stored Milo neighborhood-level results.

• fix: Preserve scCODA’s "automatic" reference-cell selection in RunscCODA() and reject unknown explicit references instead of silently using the first cell type.

• fix: PrepareDB() returns only the requested MSigDB collections after caching the other generated subsets. A cached subset loads the base MSigDB database before converting gene IDs, and falls back to the normal ID conversion when that database does not contain the requested columns.

• fix: PrepareDB() keeps nested MSigDB collections. Names such as MSigDB_M2_CGP, MSigDB_M2_CP and MSigDB_M2_CP_BIOCARTA select those subsets, parent names such as MSigDB_M2 still include every nested collection, and colon forms such as MSigDB_M2:CGP are accepted. An unknown MSigDB name reports the available collections instead of failing in complete.cases(). A cached top-level collection stays usable; requesting a nested name that is not cached rebuilds MSigDB.

• fix: RunNMF() supports the CRAN RcppML release: sparse input no longer needs Matrix attached by the caller, the attach is scoped to the call, and check_r() no longer replaces an installed CRAN build with the GitHub one.

• fix: The native Gaussian GSVA kernel follows the sparse scoring walk of the installed GSVA release (<= 2.4 and >= 2.6), so RunGSVA() and CellScoring() match backend = "r" on either engine; explicitly stored zeros are dropped before scoring.

• perf: The native Gaussian GSVA walk only evaluates the running-sum extremes at the gene-set hit positions: ~2.8x faster than the previous walk and ~50x faster than the delegated GSVA path on a 2,000 x 1,000 matrix with 400 gene sets.

• fix: RunSpatialNeighborhood() returns observed label-pair tables in a deterministic, backend-independent row order, so the cpp and reference R backends produce identical() results; a backend parity regression test is added.

• fix: PRECAST consumes the explicitly selected raw-count matrix and defaults to SCOP’s exact distance-based six-neighbor adjacency instead of interpreting image pixels as array indices or using PRECAST’s axis-restricted neighbor candidates. The matrix follows PRECAST’s column-neighbor convention. Empty/invalid spatial graphs and non-finite integration outputs fail before results are written; stored parameters include the graph builder, backend arguments and edge counts.

• fix: SpatialCellChat recognizes imported Visium HD and HD bin assays, requires explicit HD pixel-to-micron calibration and tolerance, and does not classify Visium spot/bin images as segmented cells merely because they inherit FOV.

• fix: Spatial variable feature surfaces inherit their saved assay, layer, image and coordinate columns. Backend feature IDs and p/q ranges are validated; missing BH-adjusted values use the complete supplied p-value family.

• breaking: Thread counts are named cores throughout. RunLargeVis(), RunUMAP2(), RunSmoothClust(), RunscMalignantFinder(), RunSecAct(), RunMERINGUE(), RunSpatialEcoTyper() and RunCHOIR() no longer accept the backend-specific n_threads, n_thread, ncores and n_cores argument names, and the n_threads/n_thread values that RunCIBERSORT() and the AUCell scoring path previously accepted through ... are gone as well; cores is forwarded to each backend under the name that backend expects.

• feat: FeatureDimPlot() defaults assay-gene headings to italic and supports title.face plus scalar or feature-named title.color, preserving numeric metadata headings and expression palettes.

• fix: Resolve the thisutils Python logger from scripts/, with compatibility for older installations using python/, in both reticulate imports and subprocess runners.

• fix: Dependency checks use the default check_r() behavior, allowing missing packages to be installed during analysis. Dependencies can also be installed ahead of time through PrepareEnv().

• perf: The scVelo backend builds the velocity transition, the terminal states and the pseudotime as sparse matrices instead of dense cell-by-cell ones; at 50,000 cells the velocity graph drops from about 56 GiB to 11 MiB. The terminal states and pseudotime now use iterative eigensolvers, which can return fewer regions than the previous dense implementation when the graph contains several near-closed classes, and this path requires RSpectra.

• fix: CNVPlot(plot_type = "heatmap") honours its documented heatmap_palette through thisplot::palette_colors() instead of ignoring it, and SCENICPlot() resolves its automatic reduction through DefaultReduction().

• docs: IntegrationBenchmarkPlot() documents its real workflow, and RunDeconvolution() documents the MuSiC path.

• fix: Align the C++ ports with their reference implementations: RunPalantir(backend = "cpp") uses the reference Markov-chain bandwidth and pins the start waypoint pseudotime to zero, RunCellRank(backend = "cpp") reports lineage-driver p-values from the Fisher z transform instead of a Cauchy tail, and RunSCVELO(backend = "cpp", mode = "stochastic") applies the reference 95% percentile weighting mask. Existing results for these methods change accordingly.

• fix: RunPCA() takes cores and forwards it to the native backend, which bounds its BLAS calls instead of always using the process default. The bound only applies when the linked BLAS exposes a thread-count setting (OpenBLAS or MKL); the Windows reference BLAS has none.

• fix: RunscTenifoldKnk() restores the previous OpenMP and BLAS thread counts on exit instead of leaving them changed for the rest of the session.

• fix: RunCHOIR() checks and installs its pinned backend when needed, and reports installation failures.

• fix: RunSCENIC() requires arrow for its ranking databases and reports the missing package instead of silently falling back to a rank-based approximation.

• refactor: Use the thisplot Pastel1 palette for volcano enrichment overlays instead of a hard-coded palette.

• refactor: Keep fixed method lists, resource filenames, and internal contract values at their use sites instead of package-level static constants.

• fix: Restore Palantir thread settings on exit, validate Python runtime changes before applying thread configuration, invalidate CytoTRACE2 model caches when assets change, and reject conflicting cores/ncores arguments.

• refactor: Trajectory Python backends call PrepareEnv() directly and use its managed environment validation/cache instead of bypassing preparation when modules happen to be importable.

scop 0.9.2

• fix: Cell2location stores plotting matrices for all original spots, with NA for filtered observations, and retains the modeled IDs. Historical compact matrices are expanded only when their modeled/dropped-ID records explain the missing rows; stale or unknown IDs still fail.

• fix: Long-format spatial point/jitter maps forward explicit cutoffs and quantiles without changing the default full range. Both long maps and generic dominant deconvolution maps honor combine = FALSE.

• fix: Matrix spatial plots retain explicit cutoff/quantile scales, and Cell2location dominant maps forward combine and layout controls. Constant topic/abundance panels use consistent default legend breaks.

• refactor: Long-format spatial points and jitter reuse the standard point renderer. Spatial workflow parameter validation, stage-field updates and method-name dispatch are consolidated without removing preflight or result checks.

• docs: Execute the SpatialExperiment round-trip with the correct counts layer, use merged-image auto-resolution in the PRECAST template, and render the main workflow’s empty-panel and save/reload examples.

• breaking: RunSpatialGradientFeatures() now exposes only the native C++ gradient interface. The unused compatibility parameters sample_name, platform, img_scale_fct, assay_modality, trajectory_id, width, annotation_id, core, distance, angle_span, resolution, model_add, model_subset, model_remove, and control, plus unsupported annotation_ids and ..., are removed and now fail as unused arguments. Use image/coord.cols/coordinate_space, the annotation metadata inputs, and the native screening controls documented by the function.

• fix: Spatial proportion, topic-proportion, and q05-abundance point maps share one renderer. Scale/title formatting now happens before the combine decision, so empty panels, constant values, named colors, and combined/non-combined returns have the same interpretation. Ordinary benchmark summary tables use the stable bar view for plot_type = "auto"; funkyheatmap remains explicit.

• feat: RunSpatialIntegration() gains store_object = TRUE. Setting it to FALSE omits the complete native PRECAST object while retaining the standardized result, source, parameters, summaries, and plotting payload.

• docs: Refresh the three Spatial Analysis articles around the real Visium baseline, platform-specific imports and units, and separate Seurat–SpatialExperiment/Giotto bridge contracts. External paths are templates and optional backend examples are labeled as such.

• chore: Package attachment no longer prints the ASCII startup banner or repeated environment instructions. Explicit Python-environment initialization still reports its actual success or failure.

• fix: Spatial plots preserve coordinate aspect ratios, factor order and named category colors. Point, long-format and pie facets accept metadata names with spaces, constant features honor explicit color limits, and ambiguous value IDs are rejected.

• fix: Spatial integration maps render every sample with its matching image; aligned views use aligned coordinates without raw-image overlays. SpatialCellPlot() adds named-list/layout and legend controls, with a sequential default palette for numeric features.

• fix: Deconvolution summaries report assigned/unassigned spot counts without assigning zero-weight spots to the first cell type. Invalid weights fail before object mutation. Empty deconvolution panels remain plottable and constant proportion panels share comparable legends.

• dependencies: Require thisutils >= 0.5.2 for the detect_cores() calls already used by SCOP.

• fix: Restore cores forwarding to the ncores argument in SpatialEcoTyper deconvolution and both MERINGUE R permutation tests; the deprecated ncores alias remains supported.

• cleanup: Remove the unreachable per-gene C++ branch after MERINGUE’s batch kernel returns; the batch implementation and numerical behavior are unchanged.

• fix: Forward theme_use and theme_args in all SpaTalk and COMMOT CCC plot views. Consolidate SpaTalk result preparation and remove duplicate coordinate checks and redundant plot-call wrappers.

• breaking: RunMonocle2() drops the backend = "cpp" path and its n_neighbors parameter. Monocle now accelerates the DDRTree ordering hot paths natively in C++, so the separate scop implementation is no longer needed; backend and n_neighbors are accepted with a deprecation warning and ignored. The group.by-based root_state resolution that only the C++ backend supported is removed with it, and root_state must be a trajectory State as with the previous "r" backend.

• perf: RunMonocle2() DDRTree ordering now runs on monocle’s native C++ fast paths (cell-to-MST projection, projected-cell minimum spanning tree, tree ordering traversal, and state-based root selection, ported from the former scop backend into mengxu98/monocle). Ordering a 3,000-cell dataset drops from ~9.3 s to ~1.2 s, and the ordering step no longer scales quadratically with cell number. Pseudotime, states, and the root cell are unchanged. The updated monocle is declared through check_r() and installed by PrepareEnv().

• breaking: Public APIs take the data object as object= rather than srt=. Every exported function that took srt= now takes object=; srt= remains accepted as a deprecated alias that warns on use and is removed in scop 1.0.0. RunVECTOR(), VECTORPlot(), RuntAge(), tAgePlot(), RunscTenifoldNet(), and PrepareSCExplorer() take object= only.

• breaking: Remove unverified RunSpatialDWLS() and its plotting path rather than presenting clipped least-squares output as SpatialDWLS. Use RunRCTD(), RunCARD(), or RunSPOTlight() with SpatialDeconvolutionPlot().

• breaking: Remove SpatialEcoTyperCompositionPlot() and SpatialEcoTyperSpatialPlot(). Use CellStatPlot() and SpatialSpotPlot().

• breaking: SpatialGradientPlot() now reads only stored summary, line, and model views. surface/combined plot types and current-assay/layer refetch are removed.

• fix: RunDoubletCalling() forwards the data object as object= only, so RunCellQC() no longer fails with “Provide only one of object or srt” for any db_method. The Python bridges (RunSCVELO(), RunCellRank(), RunPalantir(), RunPAGA(), RunWOT(), RunCellTypist()) no longer pass the data-object argument on to Python functions that accept neither object= nor extra keywords.

• fix: Giotto bridges reuse an already-installed GiottoClass instead of reinstalling drieslab/Giotto when the GitHub remote does not match (for example the optional CI pin giotto-suite/Giotto). The live round-trip test uses a tiny in-process fixture instead of reloading the source tree under a 120s callr timeout.

• fix: srt_to_adata() converts velocity matrices stored as extra layers of assay_x, such as spliced/unspliced in the RNA assay of a Seurat v5 object, instead of dropping them without a message. RunSCVELO() failed on such objects with Error in scVelo analysis: 'Ms', because scVelo skips the moment computation when the layers are absent and reports the problem later as adata.layers['Ms']. Names in assay_y that are neither an assay nor a layer of assay_x are now reported with the available ones, and the Python workflow names the missing layers instead of raising a KeyError.

• fix: MistyRPlot(measure =) selects the numeric result column explicitly (gain.R2/gain.RMSE or contribution/importance).

• fix: Spatial integration (PRECAST) domain, embedding, and coordinate results must match Seurat cell identities exactly, including named, reordered, missing, extra, and duplicate IDs.

• fix: SpatialCellPlot() fetches feature values through GetAssayData5() with explicit assay/layer.

• fix: StatialKontextualPlot() draws connecting lines only when a relationship has multiple radii.

• fix: Coordinate v3 accepts the empty native orientation slot introduced by newer SeuratObject versions. List integration resolves explicit image maps before merging duplicate image names. Existing coordinate-contract v2 results still require rerunning.

• fix: Spatial coordinate contract v3 uses persistent image axis provenance instead of guessing from mutable metadata, parses numeric factor coordinates correctly, and records the resolved coordinate columns. SetSpatialImageAxes() preserves custom VisiumV2 conventions through subsetting and renaming; the bundled Visium object is migrated without changing its measurements. Coordinate-dependent results from contract v2 must be rerun.

• fix: Spatial point and pie plots default to the same FOV/metadata y orientation as network and boundary plots (flip.y = FALSE). SpatialExperiment exports default to raw coordinates and retain transforms for reversible display-coordinate round trips.

• fix: Spatial integration and SpatialEcoTyper multi discovery resolve each sample’s image independently and reject incomplete image coverage. SpatialEcoTyper single discovery requires a single selected image and stores its raw coordinate source. SpotSweeper local outlier results with verified sample-prefixed IDs are restored to input order without accepting changed coordinates or expression.

• fix: Gradient result storage and VariableFeatures updates are independent; explicit empty updates clear VariableFeatures and completion messages describe actual actions.

• fix: Neighborhood spatial counts respect saved analysis scope and condition, distinguish zero from unevaluated cells, reject unsupported plot arguments, and support valid empty observed results.

• fix: RunMonocle2() skips estimateDispersions() unless feature_type = "Disp" or show_plot = TRUE. The sparse-matrix 2^31-1 crash in dispersion fitting is fixed in mengxu98/monocle 2.9.3.

• fix: RunScTenifoldKnk() builds the gene regulatory network ensemble with scop’s native C++ kernel in every backend instead of calling scTenifoldNet::makeNetworks(); backend only selects how the downstream differential-regulation steps run (r = upstream helpers, cpp = scop kernels) and what is stored, with the cpp path much faster on large gene sets.

• fix: RunScTenifoldKnk(backend = "cpp") no longer loses network scores when a subsample drops an all-zero gene: the empty sparse template was a pattern (ngCMatrix) matrix that coerced the ensemble coefficients to logical, so affected networks reached the tensor decomposition as 0/1 and the knocked-out gene was reported as not dysregulated (#427).

• fix: Optional GitHub-only backends (presto, SpatialCellChat, NNLM, SpaTalk) are resolved at their call sites through check_r() and dropped from the check-time Suggests, so R CMD check no longer installs them; presto stays a runtime-optional backend with marker fast paths using an existing installation and RunCellChat(do.fast = TRUE) resolving it before execution.

• fix: C-SIDE completion messages respect result storage and document compatibility metadata as run-level significant record counts.

• feat:

  • SpatialNeighborhoodProfile() returns observed per-cell neighbor counts and fractions at multiple raw-coordinate distances, retaining the full selected tissue context for target cells without modifying the Seurat object or storing an edge table.
  • RunDEtest() sample-level methods gain min.cells.sample (default 10) to drop pseudobulk samples whose cell count in the current group.by level is below a threshold, which can reverse DE direction relative to cell-level tests. Set min.cells.sample = 1 to keep the previous unfiltered behavior.
  • RunNichenetr() and RunMultiNichenetr() gain merged_table_file to export a temporary merged ligand-receptor/activity/target table as CSV; an existing file is never overwritten.

• spatial workflows: Add BANKSY and SmoothClust to RunStandardWorkflow(workflow = "spatial"), with explicit spatial_cluster_params. Optional SpotSweeper QC and SpaNorm normalization retain stage status and keep original counts separate from normalized downstream expression.

• spatial inputs: ReadSpatialData() delegates Visium, Visium HD and Xenium loading to Seurat and records analysis units, assay/image resolution and coordinate units. Import and workflow entry points automatically validate the selected context.

• large spatial data: ScaleData() now supports Assay5 cell subsets and updates layer membership on rescaling. SPARK-X retains sparse input, and dense spatial-feature conversions have an explicit size guard.

• normalization: Explicit do_normalization = FALSE is respected by preprocessing, including externally normalized data that the heuristic cannot classify. The SpaNorm workflow retains matched original counts for library QC alongside normalized expression.

• dependencies: Declare the existing direct lifecycle dependency used by PseudobulkExpression; no optional Remotes backend is promoted to Imports.

• data: Remove the derivative pbmc_celltypist_sub, visium_human_pancreas_results_sub, visium_human_pancreas_pair_sub, and xenium_human_pancreas_boundaries_sub datasets and one-off data migration scripts. Examples reuse pbmcmultiome_sub and visium_human_pancreas_sub, computing results when needed; specialized workflows document their required inputs without placeholder files. Existing scripts using the removed data() names must be updated.

• chore: Optional-package probes in package code now use check_r() and get_namespace_fun() instead of requireNamespace(). Documented examples call public functions under \dontrun and do not invoke check_r.

• chore: Cleared the remaining R CMD check findings across plot wrappers and integration benchmarks.

• docs: Synced the RunMonocle3() use_partition default and the RunDoubletCalling() backend default with their sources, and bolded the README headings to match the other packages.

• test: Dropped the stale AUCell namespace expectation from the consistency tests.

• removed:

  • Removed the legacy spatial registry and result-discovery APIs (GetSpatialResult(), ListSpatialMethods(), SpatialBackendStatus(), and SpatialResultInfo()). Use each producer’s documented srt@tools entry and, where provided, its dedicated result accessor or plotting function.

scop 0.9.1

• feat:
  • Internal preprocessing and integration workflows now use scop’s Seurat-compatible entry points, enabling validated native acceleration for normalization, variable-feature selection, scaling, PCA, and neighbor search while retaining transparent Seurat fallback for unsupported paths.
  • The native SCTransform(vst.flavor = "v2") path now supports validated numeric and factor vars.to.regress designs and latent.data; unsupported regression designs continue to delegate to Seurat.
  • The native PCA auto path accepts smaller stable trailing eigengaps after numerical and neighborhood-parity validation, allowing more datasets to use the accelerated implementation.
  • Added CellScoringPlot() for paired group/score UMAPs, score distributions, threshold proportions, and complete multi-program layouts from CellScoring() results.
  • CellRank + Palantir fate workflow now supports MAGIC layers, explicit Schur sizes and terminal-state validation, normalized fate/driver/transition payloads, CellRank GAM trend modules, module enrichment, and result-only CellRankPlot() views.
  • Added RunCOMMOT() and COMMOTPlot() for official COMMOT spatial communication through an isolated Python subprocess with raw-coordinate provenance and optional external H5AD storage.
  • Added RunSpaTalk() and SpaTalkPlot() for official SpaTalk single-cell and spot-reference spatial communication workflows with compact unified CCC storage.
  • Added CCCNetworkPlot(plot_type = "spatial"), a stored-result slice overlay for SpatialCellChat, SpaTalk, and COMMOT with abundance-scaled group nodes, sender-coloured directed edges, and proportion-aware composition pies.
  • RunCNV() backend coverage: copykat, fastCNV, SCEVAN, infercnv, and numbat are now verified against their original pipelines on real bundled datasets (breast-cancer scRNA-seq, infercnv example, numbat ATC2 example), with exact prediction agreement for copykat/SCEVAN/numbat and Spearman rho = 1.0 for fastCNV/infercnv.
  • The MERINGUE C++ permutation kernel now traverses the weight matrix sparsely (row-major edge list, bit-identical accumulation): nperm = 200 Moran p-values at 300 spots drop from ~442s to ~14s (~30x).
  • Removed wrappers whose upstream dependencies are unavailable or uncompilable: RunSpatialQM(), the RunSemla*() wrappers, the BASS/SpatialMNN branches of RunSpatialIntegration() (PRECAST unaffected), and the SPATA2 backend = "r" path of RunSpatialGradientFeatures() (C++ backend only; srt_to_spata2()/spata2_to_srt() removed as well).
  • RunMERINGUE() gains a backend = c("cpp", "r") option; the C++ kernels replicate MERINGUE::moranTest()/moranPermutationTest() to machine precision (including the sample.kind RNG stream). backend = "r" retains the original calls and honors moran_params.
  • Real-data consistency tests verify the wrappers against the original pipelines (CellChat, SpatialCellChat, LIANA, BayesSpace, MERINGUE, SPOTlight, SpaNorm, SpatialDWLS, RCTD, BANKSY, CARD, SpotSweeper, PRECAST).
  • DorotheaPlot() now follows the shared scop plot style: horizontal "bar"/"lollipop" comparison plots, "volcano", "heatmap" via GroupHeatmap(), "dim" (activity vs expression on embeddings), "stat" (FeatureStatPlot distributions), and "targets" (regulon-target volcano). RunDorothea() examples show FeatureDimPlot() / FeatureStatPlot() on the TF-activity assay.
  • RunIntegrationBenchmark() compares integration methods with scIB-style iLISI/cLISI, ASW, graph connectivity, and ARI/NMI scores and returns a Seurat object. Tables live in srt@tools$IntegrationBenchmark and are printed with thisplot::print_colored_table(). IntegrationBenchmarkPlot() draws box (per-cell LISI), heatmap, scatter, and UMAP views. LISIPlot() is removed; use IntegrationBenchmarkPlot(plot_type = "box").
  • SCENICPlot() / SCENICPlusPlot(): heatmap_dotplot, eregulon_dim, coverage, network, network_graph, egrn, overlap. "network" draws one hub per TF. Network plots have no title. Networks use "Chinese" when palette = "RdYlBu". Multi-TF legends show the top RSS cell type.
• fixed:
  • R-CMD-check now installs presto directly from immunogenomics/presto, avoiding dependency-resolution failures caused by treating the GitHub-only package as a CRAN package.
  • RunCellRank() now disables PETSc explicitly when retrying fate probabilities with SciPy’s direct solver, preventing CellRank 2.0.7 from silently reverting the retry to GMRES and rejecting harmless floating-point negative probabilities (#380).
  • RunCellRank() defaults to the dependency-free brandts Schur method again, so the standard scop_env no longer fails because the optional petsc4py and slepc4py packages are absent. Users can still request schur_method = "krylov" when those packages are installed (#380).
  • CellRank stored-result plotting now follows SCOP’s theme_scop, Chinese, and Spectral visual contracts; draws transition projections directly from the saved sparse transition matrix; supports random walks without requiring trend results; facets trend modules; and adds module-enrichment dot plots. CellRank enrichment selects the database gene-ID column that overlaps the actual driver background, including three-column GO tables with both Entrez and symbol identifiers.
  • Linux shared-library preloading keeps a package-local load_shared_library() seam, so dependency-order retries can be tested without attempting to mock the base dyn.load() function as a SCOP namespace binding.
  • Coordinate-contract validation now covers unified CCC plots for SpatialCellChat, COMMOT, and SpaTalk without rejecting coordinate-independent SpatialDWLS proportions. Explicit coordinate_space = "legacy_display" analysis preserves its historical lowres-to-hires fallback; current display plotting remains strict about the selected scale.
  • Spatial coordinate contract v2 now reads full-resolution VisiumV1/VisiumV2 coordinates, restores the imagerow/imagecol order that Seurat’s standard VisiumV2 loader stores positionally as centroid x/y, preserves explicit generic FOV x/y, normalizes only explicitly named legacy image row/column coordinates, and applies exactly one selected lowres or hires display transform. SpatialCoordinates(), HE-overlay point/pie/network plots, and their public wrappers gain an appended image.scale argument; variadic wrappers keep it named-only after .... Raw SpatialCellChat analysis no longer requires an available display scale, and coordinate-independent producers no longer claim a coordinate-contract version. Users no longer need to modify Seurat @scale.factors slots. Coordinate-dependent stored results created under an older or unknown contract must be rerun.
  • RunSlingshot() / CellDimPlot(lineages=) no longer fail with scale$guide == "none" on ggplot2 >= 3.5 when ggnewscale is older than 0.5.0. Colour/fill scales now keep a character guide, legend overrides go through guides(), ggnewscale is pinned to >= 0.5.2, and a failed lineage plot no longer discards the Slingshot result.
  • RunGSEA() / RunEnrichment() no longer fail with “replacement has 0 rows, data has N” when the cached database carries an empty version. The KEGG REST API dropped the Release line from /info/<org>, so PrepareDB() now resolves the KEGG version from the release-notes page when the info endpoint does not report it (retrieval-date fallback; HPO hardened the same way), and the enrichment writers tolerate missing versions.
  • RunSpaTalk() now reuses stored RunRCTD() weights through SpaTalk method 2, records the deconvolution source, and accepts a working installed SpaTalk package even when its DESCRIPTION lacks GitHub remote metadata.
  • RunLIANA() now preserves the field set returned by LIANA’s official consensus mode: rank consensus no longer gains all-NA aggregate/significance placeholders, aggregate consensus no longer gains all-NA magnitude/specificity placeholders, and deterministic ligand-receptor display identifiers are completed during SCOP/LIANA export.
  • RunCNV() attaches backend packages to the search path during execution so lazy data objects (copykat full.anno, fastCNV geneMetadata, numbat gtf_hg38) are resolvable from backend namespace code; copykat is installed from its GitHub source (navinlabcode/copykat) because it was removed from CRAN.
  • RunCNV(numbat) prefers the plain seg column over seg_label when reading joint_post output, so segment identifiers are not polluted with the (loh) suffix.
  • RunCNV() numbat extraction is vectorized (direct index assignment with duplicate averaging instead of row-by-row aggregate/strsplit, plus split()-partitioned bin statistics): a 300-seg x 2000-cell joint_post extraction drops from ~16s to ~1.3s (~12x) with bit-identical results.
  • visium_mouse_brain_slices_sub rebuilt with the current SeuratObject so bundled VisiumV2 images can be subset again.
  • RunRCTD() extracts per-spot all_weights from spacexr >= 2.2 doublet_mode = "multi" results.
  • RunSpotSweeper() detects upstream SpotSweeper::localVariance breakage on SpatialExperiment colData and degrades gracefully.
  • RunSpatialIntegration(PRECAST) writes the raw coordinates expected by CreatePRECASTObject.
  • RunLIANA() recognizes logfc_comb (and other liana method score columns) as the interaction score.
  • Exported theme_scop / theme_spatial are listed in the pkgdown reference index; the theme_scop alias test compares function equality rather than identity so installed-package checks pass.
  • DorotheaPlot(plot_type = "volcano") plots every tested TF, caps Wilcoxon underflow on the y-axis, colors points by direction, and labels significant TFs. "targets" labels only significant genes so leader lines no longer pile at x = 0.
• removed:
  • The CaSpER branch of RunCNV() is removed: runCaSpER() segfaults deterministically on R >= 4.5 (HMM manualSegment, even on 50 cells), upstream is unmaintained since 2019, and its Bioconductor dependency set is archived. RunCNV() now supports copykat, fastCNV, scevan, infercnv, and numbat.
• changed:
  • RunBenchmark() / BenchmarkPlot() are replaced by RunSpatialBenchmark() / SpatialBenchmarkPlot(). The old names are removed.
  • RunCell2fate() examples call RunStandardWorkflow() and plot time, uncertainty, module activation, and module state. env_requirements() also shows the standalone Cell2fate Python 3.9 stack.
  • ListLIANAResources() is replaced by ListCCCDB(), which enumerates ligand-receptor databases and prior models across all CCC backends with a unified db/species/status schema.
  • New PrepareCCCDB() prepares the CellTalk and CellChat databases (TERM2GENE/TERM2NAME, R.cache-backed); the CellTalk/CellChat branches were removed from PrepareDB().
  • Database listing outputs are unified (ListDB() and ListCCCDB() return plain data frames; colored console rendering is available via thisplot::print_colored_table()).
  • Python-converted object interop: PAGAPlot(), VelocityPlot()/CellDimPlot(velocity=), RunPAGA(use_rna_velocity), RunCellRank(), DE consumers (VolcanoPlot(), DEtestPlot(), RunEnrichment(), RunGSEA(), …), and RunScissor() now fall back to scanpy/scVelo result slots (paga, rank_genes_groups, velocity_umap, dpt_pseudotime, connectivities) stored by adata_to_srt().

scop 0.9.0

• feat:
  • New optional-backend wrappers: RunCHOIR() for significance-based hierarchical clustering and RunCell2fate() for RNA-velocity modeling, both with pinned backend provenance, strict result validation, metadata writeback, and resumable artifacts.
  • Added RunCell2location() for official Python cell2location reference-signature learning and spatial abundance mapping, with manifest-validated resume, signature reuse, and Cell2locationPlot() visualizations.
  • Added RunCNV() as a unified CNV workflow with optional copykat, infercnv, SCEVAN, and fastCNV backends, standardized srt@tools[["CNV"]] storage, and CNVPlot() visualizations; added RunESTIMATE() for stromal/immune/tumor-purity scoring with EstimateScorePlot() / EstimateGenePlot().
  • GRN analysis extended: RunSCENICPlus() for SCENIC+ multi-omics workflows, plus standalone RunGRNBoost2(), RunGENIE3(), and RunCisTarget() modules sharing the SCENIC regulon infrastructure.
  • Added RunSpatialCellChat() as a SpatialCellChat v3 producer for cell-, spot-, and composition-level communication, with raw-to-micron coordinate contracts and schema-v1 provenance; GetCCCObject() and SpatialCellChatPlot() provide stored-result access.
  • Spatial analysis framework: added spatial contract schema v1 (SpatialCoordinates(), GetSpatialResult(), GetSpatialGraph()), a shared spatial method registry (ListSpatialMethods(), SpatialBackendStatus(), SpatialResultInfo()), strict raw-coordinate defaults, and sparse BiocNeighbors-backed graph kernels for SmoothClust, spatial-variable-feature, and spatial-neighborhood paths.
  • Added RunSpatialBenchmark() and SpatialBenchmarkPlot() for failure-tolerant comparison of stable spatial-domain clustering methods (ARI/NMI/purity, runtime, peak memory); added ClusterTreePlot() for multi-resolution clustering trees.
  • Visualization improvements: FeatureStatPlot()/ExpressionStatPlot() gain auto_comparison; GSVAPlot() gains a mode = "diff" two-group pathway test; FeatureHeatmap(), DynamicHeatmap(), and GroupHeatmap() compact long multi-term annotations; CellCorHeatmap() gains legend.position.
  • Performance: C++ acceleration for the core Seurat flow — NormalizeData(), FindVariableFeatures(), ScaleData(), PCA and neighbor search — and for marker detection (FindMarkers(), FindAllMarkers(), FindConservedMarkers()); native kernels for manifold learning (UMAP, PaCMAP, TriMap, PHATE, DM), trajectory inference (PAGA, SCVELO, VECTOR), Z-score/PLAGE/ssGSEA/AUCell gene-set scoring, spatial-variable-feature statistics, cross-KNN and connectivity graphs, and integration embeddings.
  • RunMetabolism() now builds gene sets via PrepareDB() by default with species-aware conversion; PrepareDB() accepts local single-file database sources (MSigDB, CORUM, JASPAR, TRRUST, CellChat, and more); RunSCENIC() gained a genome parameter for hg38/hg19 reference selection.
  • RunSpatialNeighborhood() defaults to observed-neighborhood summaries; the standard spatial workflow records per-stage requested/completed/skipped/failed state and reports partial completion.
• changed:
  • Renamed standard_scop() to RunStandardWorkflow() and integration_scop() to RunIntegration() (multi-method support via integration_methods); renamed db_scDblFinder(), db_scds(), db_Scrublet(), and db_DoubletDetection() to RunscDblFinder(), Runscds(), RunScrublet(), and RunDoubletDetection(). Old entry points emit deprecation warnings and will be removed in version 1.0.0.
  • Spatial analysis, plotting, and framework conversion now require explicit image selection for multi-image objects; distance-sensitive producers default to raw acquisition coordinates (coordinate_space = "legacy_display" remains available).
• data:
  • Added visium_human_pancreas_sub, a Visium human pancreas spatial example dataset (GSE254829) with spatial image data and CODA-derived annotations, for spatial analysis.
  • Added a Xenium package-data import script for preparing curated Xenium example resources outside the installed package payload.

scop 0.8.9

• Cell-cell communication:
  • RunLIANA() now stores LIANA’s official multi-method consensus from rank_aggregate() or liana_aggregate() alongside the historical method-stacked and scop-postprocessed tables. Human and mouse defaults use the Consensus and MouseConsensus ligand-receptor resources, respectively, and multiple resources are aggregated independently.
  • Added ListLIANAResources(), ListCCCMethods(), CCCResultInfo(), and GetCCCResult() for resource discovery and Seurat-centered CCC result access.
  • RunCCC() now registers all seven SCOP CCC producers while retaining CellChat, CellPhoneDB, and LIANA as the default core run. Unified CCC plotting separates backends by default and provides explicit support-count and within-method percentile-rank views.
  • Hardened unified CCC plotting and access semantics: context filters are applied before cross-method support/rank combinations, CellChat cached rows honor the requested significance threshold, LIANA resources remain separate during export, and unified retrieval preserves backend labels. RunCCC(skip_failed = TRUE) now records preflight failures without aborting otherwise runnable methods.
• fix:
  • RunGSVA(): Removed redundant dense as.matrix() conversion in single-cell mode for backend = "r"; sparse expression matrices now stay sparse through row filtering, reducing peak memory and avoiding unnecessary dense materialization.
  • RunCellQC(): Replaced Seurat::SplitObject() with lazy cell-name-based subsetting, avoiding full-object duplication per split group. Added caching of GetAssay() gene names, GetAssayData5() counts, and per-cell QC metrics (nCount / nFeature) outside the species-check loop to eliminate repeated data extraction.
  • RunDEtest(): Pre-computed cell index mapping (split(names(cell_group), cell_group)) for paired-marker tests, replacing per-pair which() calls with O(1) list lookups.
  • AnnotateFeatures(): Replaced per-detail sapply(… "[") with type-stable vapply(…, character(1)) to avoid implicit list-to-character coercion.
  • run_scomm(): Deferred dense conversion for reference and query subsetting by removing premature as.matrix() calls; sparse matrices are subset first, then densified only when required.
  • RunUMAP2(): Replaced isSymmetric(as.matrix(graph)) with sparse-aware Matrix::isSymmetric(graph) in symmetry checks and graph subset sampling, avoiding dense materialization of large neighbor graphs.
  • RunUMAP2(): Replaced apply(as.matrix(graph), 2, order) in the uwot-predict path with thisutils::run_sparse_topk_stored(by = "col"), avoiding full dense conversion and column-wise R-level apply() while retaining graph-edge semantics.
  • RunDimsReduction() (PCA centering): Uses SeuratObject::LayerData(…, features = features) to read only HVF rows from scale.data instead of loading the full matrix followed by manual subsetting.
  • RunMDS(): Removed as.matrix() before Matrix::t(); proxyC::dist now receives the sparse matrix directly, avoiding dense conversion of large count matrices.
  • integration.R (fastMNN): Removed three as.matrix() calls passed to batchelor::fastMNN(), which accepts sparse matrices directly via SingleCellExperiment.
  • RunStandardWorkflow(): Replaced full scale.data matrix load (via GetAssayData5()) with direct layer access (SeuratObject::GetAssayData for Assay5; @scale.data for Assay) to retrieve only rownames when checking whether HVFs have been scaled, substantially reducing peak memory during the ScaleData decision step.
  • GetAssayData5.Assay(): Fixed parameter naming to use positional matching for the slot/layer argument, so GetAssayData5() correctly retrieves counts, data, and scale.data layers in both Seurat v4 (slot) and v5 (layer). Previously the named layer argument was silently dropped by SeuratObject v4, always returning the data slot regardless of the requested layer.
  • CSS_integrate(): Added Assay5 guard before SeuratObject::JoinLayers() to avoid errors with Seurat v4 Assay objects, which do not support layered storage.
  • RunDimsReduction(): Added Seurat v4 fallback for PCA centering — SeuratObject::LayerData(…, features = …) is Assay5-only; v4 Assay objects now use GetAssayData5() with manual feature subsetting.
  • RunIntegration(): Added early detection for v5-only integration methods (CCA, RPCA, fastMNN5, Harmony5, scVI5) to provide a clear, actionable error message when used with Seurat v4 Assay objects, rather than failing deep in the call stack.
• feat:
  • Added loom_to_srt() for pure-R loom-to-Seurat conversion via rhdf5, preserving velocity spliced and unspliced layers as assays without initializing Python, and added Python-backed loom_to_adata() for users who need AnnData output.
  • Added RunBulk() as a unified bulk-strategy entrypoint with method-vector selection for bulk DE, deconvolution, and cell-type-specific DE workflows.
  • Added method-specific bulk runners for de_limma_voom, de_edgeR_qlf, de_DESeq2, de_dream, deconv_MuSiC, deconv_BisqueRNA, deconv_BayesPrism, and csde_TOAST.
  • Standardized bulk results under Bulk$results$de, Bulk$results$deconv, and Bulk$results$csde, keeping DE outputs compatible with existing DEtestPlot(), RunEnrichment(), RunGSEA(), GroupHeatmap(), and FeatureHeatmap() data flows.
  • RunBulk(run_enrichment = TRUE, run_gsea = TRUE) now filters bulk DE rows with pathway thresholds and mirrors successful pathway results to Enrichment_Bulk_wilcox and GSEA_Bulk_wilcox, so EnrichmentPlot() and GSEAPlot() can read them through the standard srt@tools contract.
  • Deconvolution and CSDE bundles now record their computational engine in details; the current deconvolution runners use explicit backend = "internal" SCOP profile fitting and csde_TOAST uses explicit backend = "limma_interaction", so package backends can be wired in later without pretending to call external packages.
  • Added explicit Remotes entries for mengxu98/thisplot and mengxu98/thisutils so source installs can resolve the minimum imported versions required by SCOP. Optional bulk engines such as DESeq2 and variancePartition are installed on demand through check_r() and called lazily through get_namespace_fun().
  • Added method_args to expose method-specific tuning parameters without expanding the public API surface.

scop 0.8.8

• deps:
  • Updated the minimum dependency versions to thisplot (>= 0.3.8) and thisutils (>= 0.4.5), and removed local copies of helpers now provided upstream (thisplot::annotate_quadrants(), thisplot::clip_symmetric_range(), thisutils::collapse_sparse_rows()).
• docs:
  • Updated pkgdown reference grouping for the cell-cycle workflow.
• fix:
  • Added optional wrappers for RunDorothea(), RunBayesSpace(), and experimental RunscTenifoldKnk(). RunscTenifoldKnk() keeps the upstream scTenifoldNet workflow but fixes the QC gene-filter assignment in the local path, uses an equivalent covariance/downdate path with direct sparse matrix construction, selection-based quantile thresholding, and controlled per-gene eigensolver parallelism for large pcNet() network construction, and uses C++ helpers for tensor decomposition, manifold matrix construction, directionality, and differential-regulation distance calculations. The C++ tensor-decomposition path now computes MTTKRP updates directly instead of materializing four dense unfolding matrices.
  • PrepareEnv(): Added explicit support for mamba and micromamba executables in addition to conda, including command-name resolution, automatic package-managed micromamba download when conda = "micromamba" is requested and the command is not on PATH, manager-aware logging, micromamba env path detection, and ToS handling that only runs for standard conda. Package-managed micromamba environments now use a no-space cache root to avoid micromamba run -p path parsing failures. PrepareEnv() now also stops early if reticulate has already initialized a different Python executable, avoiding unsafe in-session switches between conda and micromamba. The Python stack now also pins setuptools < 81 so legacy packages such as trimap can still import pkg_resources.
  • PrepareEnv(): The default modules = NULL environment no longer includes scomm; install it explicitly with modules = "scomm" because the TensorFlow/scOMM stack conflicts with the default JAX/scVI stack through incompatible ml-dtypes requirements.
  • CheckDataList() / RunStandardWorkflow(): Removed an unnecessary all-feature ScaleData() call immediately after LogNormalize. Downstream workflows still scale the selected HVFs before PCA, but large raw-count inputs no longer create a dense all-gene scale.data intermediate during checking.
  • srt_reorder(): Replaced per-cluster repeated sparse subsetting and rowMeans() with a single sparse group-membership matrix multiplication, speeding the cluster-reordering step used by RunStandardWorkflow() while preserving average expression values.
  • FindExpressedMarkers(): Added a sparse fold-change and detection-rate prefilter path for common sparse RNA inputs, avoiding dense all-feature expression materialization before marker filtering while preserving default results. Supplied features are now intersected with the active layer before dense fallback scoring, which also keeps scale.data and partial-feature layers from passing absent rows into fold-change calculation.
  • C++-accelerated backends are now the default where available: RunMetabolism(), RunGSVA(), CellScoring(), RunDynamicEnrichment(), and RunEnrichment() now prefer backend = "cpp" while retaining backend = "r" for exact legacy/package behavior. RunPermutation() now uses its validated C++ implementation directly.
  • RunMetabolism() and RunGSVA(): Added a reusable C++ gene-set scoring backend for ssGSEA, zscore, and plage; RunGSVA() can now use backend = "cpp" for method = "ssgsea", method = "zscore", method = "plage", and Gaussian- or Poisson-kernel method = "gsva". PLAGE scores are oriented by the gene set mean z-score to avoid backend-dependent sign flips.
  • RunMetabolism() and RunGSVA(): Added cpp_chunk_size for the C++ GSVA kernel paths to reduce peak dense intermediate memory on large cell counts; NULL now auto-selects a chunk size for large matrices.
  • RunEnrichment(): Added an experimental backend = "cpp" ORA path using a C++ hypergeometric implementation for faster enrichment tables while keeping backend = "r" available as the clusterProfiler-compatible path.
  • CellScoring(): Added experimental backend = "cpp" support for Seurat module scoring by keeping control-gene sampling in R and moving sparse mean calculations to C++ code.
  • RunPermutation(): Uses the C++ permutation and bootstrap loops directly after validating they match the legacy R calculation for observed fractions and log2 fold-differences while running substantially faster.
  • RunUMAP2(): Added an internal C++ sparse column top-k helper for Graph inputs to speed extraction of precomputed neighbor indices/connectivities before calling uwot.
  • RunKNNMap() and RunKNNPredict(): Added an internal C++ dense column top-k helper for the raw KNN fallback to speed nearest-neighbor extraction from precomputed distance matrices.
  • PseudotimeProjectionPlot(): Reused the internal dense top-k helper for pseudotime KNN/gradient neighbor extraction from distance matrices.
  • Mapping/integration metrics: Added an internal C++ contingency-table helper for metric_accuracy(), metric_macro_f1(), metric_purity(), metric_nmi(), metric_ari(), metric_weighted_recall(), and collect_mapping_metrics().
  • RunMetabolism(): Added a C++ backend for method = "GSVA" that keeps the Poisson-kernel GSVA scoring path but accelerates repeated count-kernel calculations.
  • RunMetabolism() and CellScoring(): Added an experimental C++ backend for AUCell scoring selected only through backend = "cpp", while keeping the original R/AUCell implementation available via backend = "r" for exact package output. RunDynamicEnrichment() now passes this backend through to AUCell-based scoring.
  • CellScoring(method = "AUCell"): Fixed score-name assignment when only one feature list is scored, avoiding vector dropping before metadata column names are applied.
  • RunMetabolism(method = "VISION"): Fixed signature construction so in-memory gene sets are passed as VISION::Signature objects instead of being interpreted as file paths.
  • DynamicHeatmap() / custom enrichment workflows: Fixed incorrect term label display when user-provided TERM2GENE / TERM2NAME use non-standard column names such as term / name. Custom term annotations are now normalized consistently so that heatmap term labels show readable term names instead of fallback IDs (e.g. GO IDs). Related issue #160 (@Pineapple-wen6, @mengxu98).
  • Refactored repeated custom database input handling by introducing shared internal helpers for normalizing and assembling user-provided TERM2GENE / TERM2NAME, and applied them across RunEnrichment(), RunDynamicEnrichment(), RunGSEA(), RunGSVA(), and PrepareDB() to keep behavior consistent.
  • RunIntegration(): Improved ChromatinAssay handling by standardizing ATAC reductions after integration, avoiding non-interactive small-batch blocking, clipping TFIDF/rlsi anchor dims to available cells, auto-switching Harmony5 to legacy Harmony, and rejecting unsupported Seurat / RPCA ATAC paths with explicit messages.

scop 0.8.7

• feat:
  • Added RunDimsEstimate() and DimsEstimatePlot() for intrinsic dimensionality estimation from reductions, integrated into RunDimsReduction() and RunStandardWorkflow() to automatically select useful dimensions when linear_reduction_dims_use = NULL.
  • Renamed RunDimReduction() to RunDimsReduction() and updated downstream callers/documentation accordingly.
  • RunStandardWorkflow(): When linear_reduction_dims_use = NULL, now uses estimated dimensions stored in the reduction (via RunDimsEstimate()) when available, with fallback to the first 50 dimensions.
  • Added Seurat v5 integration methods: CCA_integrate(), RPCA_integrate(), fastMNN5_integrate(), Harmony5_integrate(), and scVI5_integrate() via Seurat::IntegrateLayers(), and exposed them through RunIntegration().
  • Added Coralysis_integrate() and exposed Coralysis through RunIntegration(), following the official Seurat v5 compatible workflow via SingleCellExperiment.
  • Added h5ad_to_srt() for reading .h5ad files directly into Seurat objects via scanpy.read_h5ad(), with automatic CSR/float64 coercion to avoid reticulate conversion issues. Layers that fail conversion are gracefully skipped and reported.
  • adata_to_srt(): Improved robustness of layer conversion — each layer is now wrapped in tryCatch() so that individual failures no longer abort the entire conversion; skipped layers are reported as warnings.
  • RunCellChat(): Enhanced to support condition-specific analyses and pairwise merged comparisons via group_column and group_cmp parameters.
  • Added RunCellphoneDB() for running CellphoneDB cell-cell communication analysis on a Seurat object through the official Python package, with support for species conversion and results stored in srt@tools[["CellphoneDB"]].
  • Added RunNichenetr() and RunMultiNichenetr() for running NicheNet and MultiNicheNet analysis on Seurat objects with standardized result storage.
  • Added unified cell-cell communication plotting functions CCCStatPlot(), CCCHeatmap(), and CCCNetworkPlot() for CellChat, CellphoneDB, NicheNet, and MultiNicheNet results.
  • Remove CellChatPlot().
• fix:
  • RunUMAP2(): Fixed reduction lookup to check existing reduction names before falling back to DefaultReduction(), avoiding errors when the exact reduction name is already present.

scop 0.8.6

• feat:
  • Added RunGSVA() for gene set variation analysis and GSVAPlot() for visualization of GSVA results. Related issue #146 (@mengxu98).
  • RunMetabolism()
    • Added RunMetabolism() for single-cell metabolism pathway scoring with support for AUCell, GSVA, ssGSEA, and optional VISION, and MetabolismPlot() for visualization
    • RunMetabolism() now uses scMetabolism KEGG / Reactome metabolism pathway definitions to identify metabolism-related terms, then rebuilds pathway gene sets from updated PrepareDB() annotations, enabling cached database updates and cross-species metabolism scoring. Related issue #146 (@mengxu98).
  • Added RunDecontX() and integrated optional decontX ambient RNA decontamination into RunCellQC(), including contamination metadata, optional decontaminated assay output, and threshold-based QC filtering. Related issue #147 (@mengxu98).
• fix:
  • FeatureStatPlot(): Fixed duplicated X/Y axis titles and main title when stack = TRUE with theme_args or title. Stack assembly now strips axis and plot titles from non-edge panels and draws a single shared ylab/title via gtable; theme styling from theme_args (e.g. axis.title.y, title/plot.title) is applied to these shared grobs. Related issue #145 (@PanSX-Dr).
• docs:
  • Updated the README badge/logo.
• data:
  • Removed the example datasets ifnb_sub, ref_scHCL, and ref_scZCL.

scop 0.8.5

• feat:
  • Default palette changed from "Paired" to "Chinese". See thisplot::ChineseColors() for details.
  • Unified the cores parameter across multiple functions (DynamicHeatmap(), FeatureHeatmap(), GroupHeatmap(), and heatmap_enrichment()), ensuring cores is correctly threaded through to RunEnrichment().
  • RunMonocle2() and RunMonocle3(): Added xlab and ylab parameters, passed through to internal CellDimPlot() and FeatureDimPlot() calls.
  • ListDB(): Now supports multiple species in the species parameter simultaneously and adds Species and DB columns to the output data frame for clearer identification.
  • Moved is_outlier to thisutils::is_outlier().
  • configure_apple_silicon_env() (Python function): Added OpenMP compatibility handling on macOS arm64 by prepending environment lib paths to DYLD_FALLBACK_LIBRARY_PATH/DYLD_LIBRARY_PATH and preloading libomp.dylib/libiomp5.dylib via ctypes.CDLL(..., RTLD_GLOBAL) to reduce scanpy/python-igraph import conflicts.
  • scVI_integrate(): Unified parameter naming by renaming num_threads to cores for consistency across integration functions.
  • Added find_neighbors_and_clusters() and run_nonlinear_reduction() helper functions to unify operations across integration functions and reduce redundant code.
• fix:
  • PrepareDB():
    • TF database now uses AnimalTFDB4 as the data source.
    • MP database - the Web Archive URL year is now dynamically determined from the current date, and the archive snapshot date for all MP-related file downloads (VOC_MammalianPhenotype.rpt, MGI_Gene_Model_Coord.rpt, MGI_GenePheno.rpt) is extracted from the server index page to ensure consistent versioning.
    • hTFtarget data download URL has been changed from "http://bioinfo.life.hust.edu.cn/static/hTFtarget/file_download/tf-target-infomation.txt" to "https://guolab.wchscu.cn/static/hTFtarget/file_download/tf-target-infomation.txt".
    • CSPA data download URL has been changed from "https://wlab.ethz.ch/cspa/data/S1_File.xlsx" to "https://raw.githubusercontent.com/mengxu98/CSPA/main/S1_File.xlsx".
    Related issus #76 (@hwa2Hu), #139 (@pengding774-dot), #140 (@mengxu98).
  • LIGER_integrate()
    • Migrated to the rliger 2.x workflow (rliger::runIntegration() + rliger::quantileNorm() on Seurat object) and now prepares/uses the ligerScaleData layer via rliger::scaleNotCenter() before integration.
    • Updated argument naming from LIGER_dims_use to liger_dims_use, and removed legacy quantile-normalization parameter compatibility mapping (ref_dataset), keeping reference as the supported interface.
  • Optimized the installation of some Python packages on Apple Silicon devices.
• docs:
  • Updated README example code and visualizations. After regenerating figures, the package size was reduced by ~3 MB.

scop 0.8.4

• fix:
  • FeatureStatPlot() / ExpressionStatPlot(): Fixed box and violin x-axis misalignment when add_box = TRUE with split.by. Groups with fewer than 2 observations are now filtered before violin density estimation (with a warning), and the violin layer uses a consistent position_dodge(width = 0.9) to match the boxplot. Related issue #123 (@oranges7).

scop 0.8.3

• feat:
  • RunDynamicFeatures(): Added PreTSA method for dynamic feature fitting. The PreTSA algorithm in scop has been re implemented to support parallelization for higher performance. Original research: PreTSA: computationally efficient modeling of temporal and spatial gene expression patterns. Use fit_method = "pretsa" for B-spline-based piecewise truncated spline analysis; fit_method = "gam" (default) keeps generalized additive models. PreTSA supports knot (0 or "auto") and max_knot_allowed when knot = "auto". Relate issue #133.
  • CellDimPlot() and FeatureDimPlot(): Added legend.title parameter (default NULL) to control the legend title. When NULL, default titles are used (e.g. group name for CellDimPlot, feature or empty for FeatureDimPlot).
  • ExpressionStatPlot() and FeatureStatPlot(): Added legend.title parameter (default NULL) for single-legend plots. When NULL, the default title (e.g. keynm or feature/group name) is used. FeatureStatPlot() forwards legend.title to ExpressionStatPlot().
  • Moved StatPlot function to thisplot::StatPlot.
• fix:
  • Differential expression plots (VolcanoPlot(), DEtestManhattanPlot(), DEtestRingPlot()): added only.pos = TRUE for positive-only visualization. DEtestManhattanPlot() now keeps the cell-type track centered at y = 0 and sizes the track from the nearest point distance around zero.
  • DynamicHeatmap() / heatmap_enrichment(): Fixed incorrect db handling when using custom TERM2GENE/TERM2NAME. Enrichment results with Database = "custom" could be incorrectly filtered by default db values (e.g. "GO_BP"), causing false “No term enriched using the threshold” warnings even when enrichment succeeded. Relate issue #133 (@1228849000).

scop 0.8.2

• feat:
  • Add the DEtestPlot() function, which calls the original VolcanoPlot() and adds two plot types, Manhattan and Ring, controlled by the plot_type parameter (c("volcano", "manhattan", "ring"), default "volcano"). Add standalone functions DEtestManhattanPlot() and DEtestRingPlot() for direct use. Relate issue #121 (@ericavalentini).
  • Differential expression visualization (DEtestPlot(), VolcanoPlot(), DEtestManhattanPlot(), DEtestRingPlot()): added res parameter to accept existing DE results (data.frame). When res is provided, srt is ignored. Data processing supports: (1) group1 or cluster column for grouped plots; (2) no grouping column for a single panel; (3) gene names from row names when gene column is missing. Relate issue #129 (@mengxu98).
  • PrepareEnv(): Update version parameter to specify the Python version of the conda environment. Default is "3.11-1" on Windows and "3.10-1" on macOS and Unix. Relate issue #103 (@PanSX-Dr).
  • RunNMF(): Add the cores parameter for RunNMF() and optimize the printed message.
  • Removed the setting in Python functions that prevents drawing functions from causing R crashes.
• fix:
  • RunPalantir(): Fixed unused plot_format parameter error. The parameter is now properly excluded from arguments passed to Python functions. Relate issue #126 (@christinejay990202-dev).

scop 0.8.1

• fix:
  • FeatureHeatmap(): Fixed group_palcolor when a named vector is passed: the function previously used only the first color for all groups because group_palcolor[[1]] on a vector returns a single element. Now when group.by has length 1, a vector is automatically wrapped as a list; when within_groups = TRUE, group_palcolor is expanded in line with group_palette.
  • GroupHeatmap(): Same group_palcolor fix as FeatureHeatmap(): support for named-vector input and correct expansion when within_groups = TRUE.
• docs:
  • group_by modified to group.by to unify documentation, involving functions: RunCellRank(), RunDEtest(), VolcanoPlot(), RunEnrichment(), EnrichmentPlot(), RunGSEA(), GSEAPlot(), RunPalantir(), RunPAGA(), RunWOT(), RunSCVELO(), releated to #120.

scop 0.8.0

• feat:
  • RunMonocle2(): New function for performing Monocle2 trajectory analysis with support for various dimensionality reduction methods (DDRTree, ICA, tSNE, SimplePPT, L1-graph, SGL-tree). Uses the fixed version of monocle2 from mengxu98/monocle.
  • RunMonocle3(): New function for performing Monocle3 trajectory analysis with support for cell ordering, trajectory learning, and pseudotime computation.
  • RunCytoTRACE(): New scop implementation for running CytoTRACE 2 analysis to predict cellular potency scores and categories (Differentiated, Unipotent, Oligopotent, Multipotent, Pluripotent, Totipotent) with support for human and mouse species.
  • CytoTRACEPlot(): New function for visualizing CytoTRACE 2 analysis results.

scop 0.7.9

• docs:
  • Optimized @inheritParams usage to reduce redundant parameter definitions.
  • Unified parameter naming: renamed group_by to group.by for consistency with Seurat conventions in RunMonocle3(), RunPalantir(), RunSCVELO(), RunWOT(), CellCorHeatmap(), and RunDynamicFeatures().

scop 0.7.8

• fix:
  • RunPalantir(): Fixed unused plot_format parameter error. The parameter is now properly excluded from arguments passed to Python functions. Relate issue #114 (@Moonerss).
  • RunSCVELO(): Fixed PAGA computation error by replacing scv.tl.paga with sc.tl.paga (scanpy implementation) for better stability. The function now uses the same PAGA implementation as RunPAGA() function.
  • RunCellRank(): Fixed GPCCA Schur decomposition error by adding fallback mechanism. When brandts method fails with “subspace_angles” error, the function automatically tries krylov method. If both methods fail, it automatically switches to CFLARE estimator for more robust computation.
  • RunCellRank(): Fixed recover_dynamics error by ensuring velocity_graph and velocity_graph_neg are properly set before calling scv.tl.recover_dynamics() for latent time computation.
• feat:
  • adata_to_srt(): Removed automatic removal of “X_” prefix from dimensionality reduction names in obsm keys. The function now preserves original reduction names as they are stored in AnnData objects.
• data:
  • Reducing the size of pancreas_sub example dataset.

scop 0.7.7

• feat:
  • adata_to_srt(): Enhanced to support multiple AnnData object types including Python AnnData objects (from scanpy/reticulate), R6 AnnData objects from the anndata package (AnnDataR6), and R6 AnnData objects from the anndataR package (InMemoryAnnData). Added internal helper functions get_adata_element() and get_adata_names() for better compatibility. Relate issues #67 (@lisch7), #91 (@mengxu98) and commit91#issuecomment.
• fix:
  • RunDEtest(): Fixed error when comparing one cluster against multiple clusters using group1 and group2 parameters. Relate issue #111 (@zhaoxiaoyan9225).
  • AnnotateFeatures(): Fixed bug where the function would fail when processing GTF file annotations due to column name matching issues during data naming. The function now correctly handles column name intersections when merging annotation data.

scop 0.7.6

• feat:
  • RunDM(): Added automatic PCA-based dimensionality reduction when using many features (>1000) to speed up diffusion map computation. The npcs parameter can be used to control the number of principal components used for pre-processing.

scop 0.7.5

• fix:
  • CellScoring(): Fixed bug where the function failed to build results. Relate issue #98 (@SuperrNaruto).
• feat:
  • RunDEtest(): Fixed compatibility issue with SeuratObject 5.0.0+ by replacing deprecated Assays() slot argument with LayerData(). Relate issue #100 (@mattizecos).
  • RunDM(): Added automatic PCA-based dimensionality reduction when using many features (>1000) to speed up diffusion map computation. The npcs parameter can be used to control the number of principal components used for pre-processing.

scop 0.7.3

• feat:
  • RunCellTypist(): New function for cell type annotation using the CellTypist method.
  • CellTypistModels(): New function for downloading and managing CellTypist pre-trained models.

scop 0.7.2

• feat:
  • RunCellRank(): Performance optimizations and code improvements.

scop 0.7.1

• fix:
  • CellDimPlot(): Fixed issue where NA values appeared in labels. Relate issue #93 (@12345nkjil).

scop 0.7.0

• feat:
  • PrepareEnv(): Integrated uv as the primary Python package installer for improved installation speed.
  • check_python(): Now uses uv as the primary installation tool with pip as fallback, significantly improving package installation speed.
  • Added find_uv() and install_uv() internal functions for managing uv package manager installation and detection.

scop 0.6.6

• docs:
  • Unified documentation format across all R functions:
    • Standardized return value tags: Changed all @returns to @return for consistency.
    • Unified parameter documentation: Replaced all \code{value} with Markdown backticks `value` format.
    • Standardized default value descriptions.
    • Added @md tags: Added @md tags to all functions using Markdown syntax in documentation.
    • Enhanced cross-references: Added @seealso links to related functions where appropriate.

scop 0.6.5

• feat:
  • PrepareEnv():
    • Added comprehensive environment variable configuration to prevent crashes when calling Python functions, including setting thread limits for OMP, OPENBLAS, MKL, NUMBA, and other libraries. This improves stability on all platforms, especially Apple silicon Macs.
    • Added accept_conda_tos() function to automatically accept conda Terms of Service for required channels, improving the conda environment setup process.
    • Fixed conda Terms of Service acceptance issue in PrepareEnv(). The function now automatically accepts conda Terms of Service for required channels, eliminating the need for manual acceptance. This addresses the issue reported in #85.
  • Multiple Python-based functions (RunPAGA, RunSCVELO, RunPalantir, RunCellRank, RunWOT, RunPHATE, RunPaCMAP, RunTriMap): Enhanced message formatting and code improvements.
  • PrepareSCExplorer(): Fixed package version dependency issues with shiny and bslib compatibility. The function now properly handles bslib theme configuration to work with both shiny 1.6.0 and 1.7.0+, addressing compatibility errors reported in #87.
• fix:
  • Improved code formatting and consistency across multiple functions.
  • Enhanced Python functions in inst/python/functions.py with better error handling and message formatting.
• docs:
  • Updated documentation for multiple functions to reflect code improvements.

scop 0.6.2

• feat:
  • CellChatPlot(): Adjusted the size of saved figures for better file size optimization.
• docs:
  • Updated README.md to remove references to Monocle2 and Monocle3 (deprecated functions).

scop 0.6.1

• feat:
  • PrepareEnv(): Improved message formatting and simplified log output for better user experience.
  • Added get_conda_envs_dir() helper function to centralize conda environment directory retrieval.
  • RunIntegration(): Enhanced integration_method parameter definition with explicit method list for better code clarity.
• fix:
  • Moved exist_python_pkgs() function to check_package.R for better code organization.
  • Replaced direct conda_info()$envs_dirs[1] calls with get_conda_envs_dir() helper function for consistency.
  • RunSCExplorer(): Updated to use thisplot::palette_list and thisplot::slim_data() instead of scop::palette_list and scop::slim_data().
  • Added thisplot to dependency checks in RunSCExplorer().
• docs:
  • Updated documentation across multiple functions.

scop 0.6.0

• feat:
  • PrepareEnv(): Enhanced with environment caching mechanism to avoid redundant environment preparation. Improved message formatting and error handling.
  • Python-based functions (RunPAGA(), RunSCVELO(), RunPalantir(), RunCellRank(), RunWOT()) now automatically call PrepareEnv() internally, eliminating the need for users to manually prepare the Python environment before using these functions.
  • cluster_within_group2(): New function for clustering within groups.
  • Multiple plotting functions: Replaced geom_sankey() with ggsankey::geom_sankey() for better Sankey diagram support.
  • Multiple functions: Replaced ::: operator with get_namespace_fun() for safer namespace access.
• fix:
  • Removed RunMonocle() function and related documentation (RunMonocle2.Rd, RunMonocle3.Rd).
  • Removed projection_functions.R file (functions moved to other locations).
  • Replaced custom theme functions with thisplot::theme_this() (exported as theme_scop()).
  • Replaced direct log_message() calls with thisutils::log_message() for consistency.
  • Removed palette_list data object.
• deps:
  • Moved cli from Suggests to Imports for better message formatting support.
  • Added thisplot to Imports for theme and utility functions.
  • Added ggsankey to Suggests for Sankey diagram support.
  • Added remote dependencies: theislab/destiny, mengxu98/thisplot.
  • Removed unused dependencies: Biobase, BiocGenerics, concaveman, DDRTree, glmGamPoi, hexbin, monocle, png, ragg, tidyr.
• docs:
  • Updated documentation across multiple functions to reflect code refactoring.
  • Improved code organization and maintainability.

scop 0.5.5

• fix:
  • Fixed VelocityPlot() function error in plot_type = "grid" mode: replaced vectorized arrow length with fixed-length arrows (using mean length) to resolve vapply() error that occurred when grid::arrow() received a vector instead of a single value, see #74.

scop 0.5.4

• fix:
  • Fixed parameter name error in CheckDataType() function calls: changed data parameter to object in RunKNNMap(), RunSymphonyMap(), RunScmap(), RunPCAMap(), and RunSingleR() functions, see #68.
  • Fixed SingleCellExperiment object creation in RunScmap() and RunSingleR() functions: changed from coercing SummarizedExperiment to directly constructing SingleCellExperiment objects.

scop 0.5.3

• feat:
  • PrepareDB(): Changed default Ensembl_version parameter from 103 to NULL for more flexible version handling.
  • Added Python version log_message() for Python-based functions (RunSCVELO(), RunPAGA(), RunPalantir(), RunCellRank(), RunWOT()) and added verbose parameter inheritance and improved message formatting through cli formatting.
• fix:
  • Delete harmonizomeapi.py file.
  • Move scop_analysis.py into a single functions.py file in inst/python/ for better code organization and maintainability.
• docs:
  • Improved parameter documentation consistency.

scop 0.5.1

• docs:
  • Improved reference formatting and consistency across multiple functions.
  • Enhanced documentation clarity and readability.

scop 0.5.0

• feat:
  • RunCellChat(): New function to perform CellChat analysis for investigating cell-to-cell communication with support for human, mouse, and zebrafish species.
  • CellChatPlot(): New function to visualize CellChat analysis results with various plot types and customization options.
  • Multiple integration functions: Improved error messages and message formatting for better user experience.
• deps:
  • Added CellChat package dependency with remote repository jinworks/CellChat.
• docs:
  • Updated README.md with improved code formatting and examples.
  • Enhanced documentation for cell communication analysis functions.
  • Improved error messages and user guidance across integration functions.
• fix:
  • Removed some example figures to optimize package installation size.

scop 0.4.0

• feat:
  • RunProportionTest(): New function to perform Monte-carlo permutation test for quantifying cell proportion differences between conditions.
  • ProportionTestPlot(): New function to generate proportion test plots with customizable significance thresholds and visualization options.
  • Multiple Python-based functions: add \dontrun{} blocks for Github workfolw checking.
• docs:
  • Added comprehensive documentation for new proportion testing functions.
  • Enhanced example documentation across multiple functions.
  • Updated package documentation and examples.

scop 0.3.4

• docs:
  • Updated workflow examples and function documentation.

scop 0.3.3

• feat:
  • Multiple functions: Improved parameter documentation formatting and consistency across the package.

scop 0.3.2

• feat:
  • GetFeaturesData() and AddFeaturesData(): Enhanced argument clarity, added input validation, and standardized return values for Seurat, Assay, and Assay5 objects.
  • CellCorHeatmap():
    • Renamed parameters: query_cell_annotation → query_annotation, ref_cell_annotation → ref_annotation.
    • Improved error message formatting through cli formatting.
    • Simplified variable assignments and improved readability.
• docs:
  • Comprehensive documentation updates across multiple functions including AnnotateFeatures, CellDimPlot, CellStatPlot, FeatureStatPlot, GroupHeatmap, RunCellQC, and others.
  • Improved parameter descriptions and function clarity.

scop 0.3.1

• feat:
  • EnrichmentPlot() and GSEAPlot(): Removed conditional font face styling (face = ifelse() logic) for better text rendering consistency. Set the default value of lineheight from 0.5 to 0.7.
  • Updated check_r() function for improved package checking functionality.
  • Updated reexports functionality.
• docs:
  • Updated documentation formatting and consistency.

scop 0.3.0

• feat:
  • Fixed segmentation faults and R crashes on M-series MacBook when running Python functions.
  • RunPAGA(): Enhanced with M-series MacBook detection and automatic environment configuration.
  • RunSCVELO(): Added ARM64-specific optimizations to prevent crashes and ensure stable execution.
  • RunCellRank(): Implemented M-series compatibility with proper NUMBA configuration.
  • RunPalantir(): Added ARM64 support with single-threaded execution mode.
  • RunWOT(): Enhanced with M-series MacBook environment variable settings.
  • RunTriMap(): Added M-series MacBook compatibility for dimensionality reduction.
  • RunPaCMAP(): Implemented ARM64-specific environment configuration.
  • RunPHATE(): Added M-series MacBook support for non-linear dimensionality reduction.
  • RunCellQC(): Enhanced both scrublet and doubletdetection functions with ARM64 compatibility.
• docs:
  • Updated function documentation to reflect M-series MacBook compatibility.
  • Added technical notes about ARM64 architecture considerations.

scop 0.2.9

• fix:
  • Fix bug for RunSCVELO().
• docs:
  • Updated documentation for some functions.

scop 0.2.7

• feat:
  • Added an internal function .check_pkg_status() to check if an R package is installed.
  • Update function CheckDataType() to S4 class function.
  • Update function RunStandardWorkflow(), make it more efficient.
• data:
  • Delete lifemap data, including: lifemap_cell, lifemap_compartment and lifemap_organ.
  • Reconstructed the sample data for both panc8_sub and pancreas_sub, retaining only the basic Seurat object.

scop 0.2.6

• feat:
  • Added remove_r() function for easy remove R packages.
  • Rename function: RemovePackages() to remove_python().
  • Removed other methods of installing R packages from the check_r() function, only retaining pak::pak.
  • Delete useless import packages: BBmisc, BiocManager, covr, devtools, promises and withr.
  • Optimize the structure of _pkgdown.yml file.
• docs:
  • Updated documentation for some functions.

scop 0.2.5

• feat:
  • Rename function: palette_scop() to palette_colors().

scop 0.2.4

• feat:
  • Rename functions: check_srt_merge() to CheckDataMerge(), check_srt_list() to CheckDataList and check_data_type() to CheckDataType.

scop 0.2.2

• feat:
  • Replace all BiocParallel::bplapply() with thisutils::parallelize_fun().
• fix:
  • Fix bugs in RunSingleR().

scop 0.2.0

• feat:
  • Added remove_python() function for easy remove Python packages.
• fix:
  • Corrected an issue in py_to_r2() function (intrinsic function), which ensures that Python-dependent functions like RunPAGA() and RunSCVELO() function run correctly.

scop 0.1.9

• feat:
  • Update CellScoring() and AddModuleScore2() functions. Now, new parameters cores and verbose have been added.
  • AddModuleScore2() function no longer uses the BiocParallel::bpparam() function to enable parallelization, but thisutils::parallelize_fun, and the cores parameter is used to control the number of cores in thisutils::parallelize_fun.

scop 0.1.5

• fix:
  • Fix error for RunPalantir() function, see #23.

scop 0.1.4

• feat:
  • Update .onAttach(), now .onAttach() will print more information about conda and Python.
  • Update PrepareEnv() function for easy add or update a conda environments and install Python packages.
  • Added ListEnv() and RemoveEnv() functions for easy management of conda environment and Python packages.

scop 0.1.3

• feat:
  • Added TACSPlot() function for creating FACS-like plots. Please refer to Kernfeld et al. paper and Github for specific information.

scop 0.0.9

• fix:
  • Fix a bug for PrepareEnv() function.
  • The default Python version is now set to 3.10-1.

scop 0.0.6

• feat:
  • Add GetAssayData5() function, a reimplementation of GetAssayData(), for compatibility with Seurat v5 Assay objects.
  • Updated GetFeaturesData() and AddFeaturesData() function to support retrieving and adding feature metadata.

scop 0.0.5

• data:
  • Updated the pancreas_sub and panc8_sub test datasets to the Seurat v5 object format.

scop 0.0.1

• Initial version