NEWS

---

PscanR 0.99.3

• Promoters of different lengths are not comparable, because a promoter's best motif score depends on its length. pscan(), pscan_filtered() and ps_build_bg() now keep the promoters of the most common length (before, those of the longest one), and the warning reports how many promoters were kept, their length, and how many were discarded.
• pscan_filtered() removes duplicated promoters and promoters of other lengths, as pscan() does.
• ps_scan() warns when it is called directly on promoters of different lengths, which it scans one at a time.
• pscan(), pscan_filtered() and ps_build_bg() stop with an error that names the motifs when the promoters kept are shorter than any motif in the collection. Before, pscan() failed with an unrelated z-test error, pscan_filtered() with "logical subscript contains NAs", and ps_build_bg() silently stored NaN background statistics. ps_scan() stops with a clear error when no promoter is long enough for its motif.
• pscan_filtered() encodes the promoters once for all motifs, as pscan() does.
• Argument checks are shared by all exported functions and stop with one wording, such as "'x' must be a DNAStringSet", without showing internal function names. Checks that were missing are added: ps_retrieve_bg_from_file() requires a file path, ps_build_bg_from_table() a data.frame, ps_build_bg() a logical fullBG, and pscan_filtered() a numeric n. ps_write_bg_to_file() reports an output file or directory it cannot write to; the check existed but never ran. ps_motif_barplot() accepts an FDR only between 0 and 1, as ps_results_table() does.
• Removed internal code without callers: .ps_natural_order(), the stored PSMatrixList class generator, and the .PS_PSEUDOCOUNT() and all_sequences_id() accessor generics (the latter read a slot that PSMatrix does not have).
• Tests cover the per-sequence scanner used for promoters of different lengths.

PscanR 0.99.2

• Changes from the Bioconductor review:
  • ps_motif_barplot() and the density and hit-score plots use the Okabe-Ito colour-blind-safe palette from grDevices::palette.colors().
  • The vignettes locate their bundled files with system.file(), so every displayed chunk can be run as shown.
  • The funder is listed in Authors@R.
  • PscanRBackgrounds, the ExperimentHub package with the precomputed backgrounds, is in Suggests now that it is part of Bioconductor.
  • Minor code cleanup: no comparisons with TRUE, startsWith() for prefix tests, toString() for comma-separated lists.
  • The code follows a lintr configuration (.lintr: 4-space indentation, 80 columns) that CI enforces; the tests were reindented to match.
  • Consistent argument names: ps_retrieve_bg(jaspar_release = ) (was JASPAR_matrix) and ps_scan(use_full_bg = ) (was use_full_BG).

PscanR 0.99.0

• Initial Bioconductor submission.
• pscan() implements the Pscan method (Zambelli, Pesole and Pavesi, 2009, doi:10.1093/nar/gkp464): it tests a set of promoter sequences for transcription factor binding motifs that score higher than in a background of all promoters of the organism, and reports z-scores, p-values and FDR.
• Precomputed backgrounds for the JASPAR 2020, 2022 and 2024 CORE collections, seven genome assemblies and five promoter windows are retrieved from ExperimentHub (package PscanRBackgrounds), or directly from their Zenodo records, with ps_retrieve_bg() and listed with ps_available_bg().
• Custom backgrounds can be built with ps_build_bg() and saved and reloaded with ps_write_bg_to_file() and ps_retrieve_bg_from_file(). Full backgrounds keep per-promoter hits, so that pscan_full_bg() can test a set of transcripts without rescanning sequences.
• pscan_filtered() restricts an analysis to motifs that pass a first scan.
• ps_select_promoters() and ps_selection_summary() choose one promoter per gene from a gene list.
• Results are summarised with ps_results_table() and visualised with ps_zscore_heatmap(), ps_motif_barplot(), ps_density_plot(), ps_hitpos_map() and ps_hit_score_plot().
• Motif scanning can run in parallel through BiocParallel; it is serial by default.