NEWS.md

In splikit: Analysing RNA Splicing in Single-Cell RNA Sequencing Data

splikit 2.3.1

CRAN Compliance

• Fixed no visible binding for '..keep' NOTE in .gene_exons(): replaced exons[, ..keep] with exons[, keep, with = FALSE].
• Fixed no visible binding for 'x' NOTE in .build_plot(): added x to globalVariables() in globals.R.
• Wrapped get_pseudo_correlation() example in \donttest{} to avoid exceeding the CRAN 5-second example time limit.

splikit 2.3.0

New Features

• RefSeq GTF support in plotting functions. plot_exclusive_junctions() and plot_exclusive_junctions_event() now accept RefSeq-style GTFs that lack gene_name / transcript_name attributes. The internal attribute extractor tries gene_name -> gene_id -> gene, and transcript_name -> transcript_id, so a RefSeq annotation like NCBI's GCF_*_GRCh38 file works with no manual preprocessing.
• Structured return value. Both plotting functions now return an S3 object of class splikit_junction_plot with components $plot (ggplot), $info (tidy per-junction data.table), plus the underlying exons / junctions / tx_order tables. The info data.table includes gene_name, gene_id, transcript_name, transcript_id, chr, strand, j_start, j_end, j_width, exclusive, n_tx_with_junction, observed_in_eventdata, row_names_mtx, and is_annot for straightforward downstream analysis.
• S3 print() method for the new class: rendering the plot and printing info when the object is auto-printed at the REPL.

Internals

• Extracted shared helpers (.gtf_attr, .load_gene_gtf, .gene_exons, .tx_junctions, .order_transcripts, .build_plot, .build_info) so the GTF-only and eventdata-based plots now share a single implementation path.

splikit 2.2.1

CRAN Compliance

• Addressed CRAN reviewer feedback on the 2.0.0 submission: replaced the un-suppressable print(summary_table) calls in make_junction_ab(), make_gene_count(), and make_velo_count() with message(paste(capture.output(...), collapse = "\n")). The summary now respects suppressMessages() and stays behind the verbose = TRUE guard. A stray if (verbose) cat(...) in make_m1() was likewise converted to message() for consistency.
• Title and Description already had the redundant "A toolkit designed for" phrasing removed in 2.0.1; no further action needed there.

splikit 2.2.0

New Features

• plot_exclusive_junctions_event() - sibling of plot_exclusive_junctions() that sources the drawn arcs from a splikit eventdata table. Exon structure still comes from the GTF and exclusivity is computed gene-wide from the annotation, but only junctions whose (i.start, i.end) coordinates are observed in eventdata are drawn. If eventdata lacks a gene_name column the function runs make_eventdata_plus() internally using the supplied GTF path. The plot is rendered directly to the current graphics device (no out_file).

splikit 2.1.0

New Features

• Transcript-exclusive junction plotting
• plot_exclusive_junctions() renders a gene model with one row per transcript, exon rectangles, intron/junction arcs, and Ensembl ids shown beneath transcript names on the y-axis. Junctions used by a single transcript of the gene are highlighted as solid black arcs; shared junctions render as thin grey arcs.
• Flags: show_exclusive (restrict to exclusive-owning transcripts), transcript (pin the plot to one or more transcript names), curvature (arc-height knob).
• plot_exclusive_junctions_pdf() writes a multi-page PDF: page 1 is the full gene view, subsequent pages show each exclusive transcript with its exclusive junction in black.
• Accepts a GTF path or an already-loaded data.table. ggplot2 is declared under Suggests.

Bug Fixes

• make_eventdata_plus() gene_name extraction
• The previous regex used sub('.*gene_name "([^"]+)".*', ...), which returns the input unchanged when no match is found. Gene records without a gene_name attribute (e.g. ~310 rows in Ensembl mouse GRCm39.110) therefore ended up with a gene_name equal to the full attribute blob.
• Fixed with a grepl()-guarded extractor that returns NA on no-match, with a fallback to gene_id so downstream matching by name keeps working. The function now emits a message() reporting how many records required the fallback.

splikit 1.0.5

🚀 Major Enhancements

Performance Improvements

• Multi-threading Support: Added parallel processing capabilities for gene expression analysis
• find_variable_genes() now supports n_threads parameter for faster computation with method = "sum_deviance"
• Consistent multi-threading implementation across all feature selection functions

• Significant speedup for large-scale datasets

• Intelligent Memory Management: Enhanced memory efficiency for large datasets

• make_m2() now includes automatic detection of memory limits
• Smart switching between fast and batched processing modes
• New parameters: batch_size, memory_threshold, force_fast for fine-tuned control
• Prevents 32-bit integer overflow issues with large matrices

New Features

• Enhanced Pseudo-correlation Analysis
• get_pseudo_correlation() now supports both Cox-Snell and Nagelkerke R² metrics
• Added metric parameter for selecting correlation type (default: "CoxSnell")
• Full support for both sparse and dense matrix inputs for M1/M2 matrices

• Improved numerical stability with singular matrix handling

• Improved Data Processing

• make_junction_ab() enhanced with:
  • verbose parameter for detailed progress tracking
  • keep_multi_mapped_junctions option for including multi-mapped reads
  • Better memory management for large junction datasets
• make_m1() now includes:
  • min_counts parameter for filtering low-count events
  • verbose option for progress monitoring
  • Improved handling of sparse events

Developer Experience

• Enhanced Logging: All major functions now support verbose output
• Detailed progress tracking for long-running operations
• Memory usage reporting in make_m2()

• Informative messages even with verbose = FALSE for critical steps

• Documentation Website: Added pkgdown support

• Automatic documentation generation via GitHub Actions
• Bootstrap 5 modern interface
• Complete function reference with examples
• Available at: https://arshammik.github.io/splikit/

🐛 Bug Fixes

• Fixed non-ASCII characters in verbose output messages (replaced with ASCII alternatives)
• Corrected parameter naming inconsistency (multithread → multi_thread)
• Removed unused Nagelkerke R² calculation in pseudo-correlation function
• Fixed gfortran dependency issues in Makevars configuration

📚 Documentation Updates

• Comprehensive updates to all function documentation with new parameters
• Added detailed vignettes:
• Complete splikit manual for single-cell splicing analysis
• STARsolo processing guide with step-by-step instructions
• Updated README with clearer examples and installation instructions
• Consistent Roxygen2 documentation across all functions

🔧 Technical Improvements

• Updated to version 1.0.5
• Improved C++ implementations for better performance
• Better error handling and input validation
• Memory-efficient sparse matrix operations
• Cleaned up package structure for CRAN submission readiness

Breaking Changes

None - all changes are backward compatible. Existing code will continue to work with default parameters.

Acknowledgments

This release includes contributions addressing issue #16, focusing on performance enhancements and multi-threading support. Special thanks to all users who provided feedback and testing.

Installation

# Install from GitHub
devtools::install_github("Arshammik/splikit")

# Load the package
library(splikit)

What's Next

• S4 class implementation for object-oriented interface (planned for 2.0.0)
• Additional statistical methods for splicing analysis
• Enhanced visualization capabilities
• Further performance optimizations

For questions or issues, please visit: https://github.com/Arshammik/splikit/issues

splikit documentation built on May 13, 2026, 9:08 a.m.