NEWS

---

SpliceImpactR 1.1.4

Bug fixes

• BioMart queries for release go to https://e<release>.ensembl.org, which Ensembl forwards to that release's archive, instead of to the archive host itself. With biomaRt before 2.70, new installations could no longer connect to archive hosts ("Unable to contact any Ensembl mirror"): biomaRt first downloads Ensembl's list of archives, which is no longer available. Release 116 now works with any biomaRt version. The data come from the same archives, so results and cached features are unchanged.
• With biomaRt before 2.70, connection errors for an archive host passed as ensembl_host explain this and suggest the e address.

SpliceImpactR 1.1.3

Breaking changes and changed results

• get_protein_features() queries Ensembl 111 by default (was 109), the Ensembl release of the default GENCODE v45 annotation from get_annotation(). Features from full runs change, and cached features are fetched again once.
• ensembl_mirror is deprecated and ignored: Ensembl retired its BioMart mirrors.

Bug fixes

• BioMart queries go to the Ensembl archive of release by its explicit host (releases 105 to 116). Ensembl 116 (June 2026) is the last release with BioMart, and www.ensembl.org no longer serves it. The new ensembl_host argument selects another host. Release 116 needs biomaRt 2.70 or later (earlier versions send its queries to www.ensembl.org), and connection errors name the host.

SpliceImpactR 1.1.2

Breaking changes and changed results

• get_splicing_impact() selects transcript pairs with the ORF-aware matcher (matching = "orf") by default; matching = "legacy" restores the previous matcher. The ORF matcher checks each form's exact event structure, never pairs a transcript with itself, labels pairs without a structural match as approximate, and returns unresolved comparisons in matching. Selected transcripts and pair counts change, and matching is slower.

New features

• get_ranked_pairs(return_class = "S4") stores the selected pairs, form rows and diagnostics in a SpliceImpactResult, where get_splicing_impact() stores them, so the stepwise S4 workflow uses the ORF matcher. The default still returns the list. get_pairs() S4 output sets metadata$matching = "legacy" and drops earlier ORF diagnostics.

Documentation

• The README and vignettes describe the ORF matcher as the default, including in the stepwise S4 workflow. The README also covers one-based rMATS coordinates, the domain-enrichment population, gene universes, new output columns and input requirements. The matching vignette notes that diagnostics are empty tables, not NULL, when no forms are significant.

SpliceImpactR 1.1.1

Breaking changes and changed results

• rMATS coordinates are converted to one-based, closed intervals: starts move by +1 and retained introns exclude the flanking exon ends. Differential inclusion tables saved with earlier versions miss exact matches by 1 nt and must be regenerated; get_rmats_post_di() returns package-format tables unchanged.
• A3SS and A5SS forms include the partner (flanking) exon in inc, in genomic order. inc strings and event counts change, and events that differ only in their partner exon are no longer merged.
• get_rmats_post_di() numbers events across files and stops when one event type appears under more than one grp1/grp2 comparison. The new case_group argument selects the case group (default 1, rMATS --b1, as before), and each import reports which group it used.
• get_background() defaults to source = "annotated" (was "hit_index"). Passing input without source warns that it is ignored; "hit_index" or "user-given" without input is an error. The gene_universe and feature_gene_universe attributes give the eligible gene universes.
• Domain enrichment counts only pairs with a tested domain change, in both foreground and background, and counts repeated transcript pairs once. Foreground pairs and their domains must be in the background, and the odds ratio compares the foreground with the rest of the background. domain_col_fg and domain_col_bg are honoured, and delim defaults to "\\|". Enrichment results change.
• The legacy matcher ranks candidates by structural fit (exon class, overlap, width) before protein-coding status, with transcript support level as the last tie-break, and candidate exons must belong to the event's gene. Selected transcripts can change.
• Forms without a matched transcript no longer produce pairs; they were paired with an NA transcript.
• summary_classification has a new "NMD" class, with top precedence, when either pair member is annotated as nonsense-mediated decay.
• import_di_table() stops on strands other than +/- and uses event_id, form and padj columns when present.
• get_rmats_hit() stops on unknown event types and returns only the requested HIT index types.
• get_user_data_post_di() requires event_id and stops on events without INC and EXC forms or a SITE form, or that mix them. Without event_id it used to give each row its own ID, so forms could never pair.
• Protein-feature cache keys hash the full input content, so features are fetched again once; old cache files are left on disk.
• Requires R (>= 4.5.0).

New features

• Opt-in ORF-aware transcript-pair matching: get_ranked_pairs() and get_splicing_impact(matching = "orf"), with matching_max_candidates and matching_fallback. Pairs carry matching_tier and ranking diagnostics, and unresolved comparisons are returned with reasons. The legacy matcher remains the default. The new transcript_pair_ranking vignette describes the protocol.
• get_annotation(filter_tsl = NULL) keeps transcripts of every support level, including missing ones, in a separate processed cache. The default remains TSL 1-3.
• keep_sig_pairs() adds site_significant, and pairs carry site_significant_case, site_significant_control and n_event_comparisons for events with several sites.
• Matched and paired tables carry each transcript's transcript_type.
• get_matched_events_chunked() gains verbose; chunk progress is reported with message().
• get_manual_features() places rows given only ensembl_peptide_id and reports rows it cannot place.
• Empty analyses return zero-row tables with their usual columns, in table and S4 modes, and plots return placeholders.

Bug fixes

• get_pairs(source = "paired") failed whenever an event had two forms; it now takes positive delta PSI as the case, like "multi".
• PPI switches use the side of each interaction the event gene is on. Domain-motif changes with the motif on gene A were never detected, and domain-domain changes could be called from a partner's domain.
• HIT index import reads plain and .gz files once (PSI was halved), finds .gz PSI and .exon.gz files, and uses anchored file patterns. rMATS .txt.gz files are read, and one-base segments are kept.
• get_rmats_hit() binds rMATS and HIT index tables by column name.
• User-given size factors ('user-given') failed; sample folders without a trailing slash now work.
• Frame checks follow transcript order upstream of the event, A5SS frame checks use the variable exon, and event lengths sum the selected transcript's exons.
• Manual features on minus-strand transcripts map to the right genomic positions, and BioMart features at identical coordinates are kept when their database or ID differ.
• plot_alignment_summary(output_file = ) saved an undefined object. Feature plots draw every interval of multi-interval inc/exc, and plot_enriched_domains_counts() counts unique events.
• get_splicing_impact() works from res alone and no longer reloads raw data when given sample_frame and res. S4 metadata is replaced, not appended, on re-runs, and has_* flags reflect content.
• Processed annotation caches are stored inside the cache folder. BioMart queries use bounded transcript-ID batches and restore options().
• Inputs are no longer modified in place.

Other changes

• Imports R.utils (>= 2.13.0), which data.table::fread() needs for .gz input.
• Removed unused internal helpers.

SpliceImpactR 0.99.4

• Version bump for Bioconductor

SpliceImpactR 0.99.2

SpliceImpactR 0.99.1

• Version bump for Bioconductor

SpliceImpactR 0.99.0

• Added a NEWS.md file to track changes to the package.