Changes in version 0.99.9 - Aligned to Python selexprep v0.4.8 (doi:10.5281/zenodo.24427025). No call, status or extracted read changes. This version is about running the package on your own reads. - selexprep_detect() warns when a called primer contains a sequencing adapter, which happens when the reads run past the construct, and prints a message when it finds one constant only, saying what the sequences will keep. - Extraction warns when both primers are set but the mode is one-sided, so that one primer is left untrimmed, and names extraction_mode = "BOTH_PRIMERS_SINGLE_READ". It warns when a supplied primer is in under 5% of the earliest round's reads, and says so when its reverse complement is there instead (a reverse primer typed as ordered). - run_selexprep() warns when the primers are split between R1 and R2 and no sequence is counted. read_selexprep_inputs() warns about FASTQ files with no reads and names a path that does not exist. - Supplied primers may carry IUPAC ambiguity codes, as the help of selexprep_extract() already said. An invalid extraction_mode lists the valid values. - The help pages give the values of extraction_mode, status and required_action, say how to write the primers, and explain why a report with supplied primers has status = "LOW". The vignette and the README have a guide for your own reads: reads longer than the construct, supplied primers, replicates, run time, and counting both read orientations. - Catalog provenance links point to the archived release (doi:10.5281/zenodo.23104868) and to Software Heritage permalinks of the source commit. The dataset and its metadata are unchanged. - Add inst/CITATION: the package, the selexprep method and cutadapt. Changes in version 0.99.8 - Credit cutadapt for the aligner ported in R/primer-match.R: its MIT license and copyright notice are in inst/licenses/cutadapt-MIT.txt, summarized in inst/COPYRIGHTS, and Marcel Martin is listed as copyright holder in DESCRIPTION. Changes in version 0.99.7 - Replace the identical synthetic-round vignette with licensed, attributed 600-read R1 subsets of three real hydrocortisone DL-SELEX rounds. Show safe inference failure, published Table S8 flanks, explicit primer overrides, extraction yields, sparse counts, QC plots and demultiplexing. - Add a vignette guide for starting from your own reads. Explain the chosen read orientation, paired-end inputs and each QC flag in the example, and export one count table per round with write_selexprep_counts(). - Show the QC retention, length and diversity plots as separate figures. Length fractions omit lengths not observed in a round, and the retention plot leaves room for its legend. - Add package-level help and runnable offline examples to all help pages. FASTQ examples create their inputs; a cached ENA inspection supports offline download planning. Only fresh ENA inspection requires a network. - Locate primers as cutadapt 5.2 does in the Python pipeline. A primer cut short at the read start (5') or end (3') is accepted when at least 3 bases align; among several occurrences the first acceptable one is kept unless a later one overlaps it or covers more of the primer with a higher score; an error at a primer edge is a substitution rather than a deletion. Synthetic edge cases and the bundled real reads are now trimmed exactly as cutadapt trims them; the example retains 228, 278 and 297 reads instead of 225, 277 and 294. - Vectorize exact primer matching and strand evidence with Biostrings; only reads with a possible alternative match are aligned. Batch the reverse-complement canonicalization of distinct QC k-mers. - Use one reverse-complement implementation for counting and extraction, fixing the internal failure on other nucleotide symbols. reverse_complement() now complements IUPAC ambiguity codes (R/Y, K/M, B/V, D/H) while keeping case, RNA input and other symbols. - Measure match rates and insert lengths for manually supplied primers on the earliest round, so the library report and the low_primer_match QC flag describe the primers actually used for extraction. - Report sides without a candidate flank as such in detection failures, instead of a 0.00 match rate. - Name reads in the opposite orientation as a possible cause in low-yield extraction warnings. - Describe the QC summaries and flags in selexprep_qc() as computed. - Simplify shared flank evidence, QC thresholds, mate pooling, barcode comparisons and verified downloads; preserve public result schemas, RNG behavior and failure diagnostics while shortening workflow functions. - Normalize lowercase RNA barcodes and reads consistently with uppercase RNA and DNA, reusing the common alphabet-normalization helper. - Document redistribution terms separately for the MIT catalog compilation, CC BY 4.0 FASTQ subsets, public metadata and synthetic regression fixtures. Preserve the Python v0.4.6 catalog values and original curation statuses; record human spot checks separately from model agreement. Changes in version 0.99.6 - Simplify ENA condition messages to address the two new BiocCheck coding practice notes, preserving four-attempt retry and sample-metadata fallback. - Keep diagnostic accessions, underlying sample failures and retry advice explicit; remove redundant failure wording from the filereport message. Changes in version 0.99.5 - Align library inference with Python selexprep v0.4.6: retain shared library constants across rounds using per-round offsets, preserve the insert boundary, and avoid ambiguous repeated-core anchors. - Warn about extraction pools retaining fewer than half their reads or half the best pool's yield, retaining diagnostics in low_yield_rounds. File workflows also retain input_yields and low_yield_inputs per FASTQ, so aggregation cannot mask a poorly matching input. - Normalize lowercase RNA consistently during detection and primer matching. - Read ENA BioSample attributes for round inference; handle conflicting and weak PCR-cycle evidence, and stop interpreting ArrayExpress RV sample labels as selection rounds. - Retry ENA filereport service-error bodies, retain run-level library strategy and sample evidence, and warn when a round pools distinct BioSamples. - Pin the 240-record catalog and provenance to Python v0.4.6, including the corrected PFOA target class. Add offline Python/R regression references. Changes in version 0.99.4 - Refuse inferred libraries with a zero-length random region instead of reporting a valid full-insert extraction. - Rescue well-supported primer cores when noisy outer flanks lower the full-length match rate, without changing the extraction boundary. - Update the bundled public catalog to the adjudicated Python v0.4.1 snapshot, retaining immutable source and evidence provenance. Changes in version 0.99.3 - Write LibraryReport and manifest JSON with UTF-8 LF line endings on every platform, preserving deterministic hashes and Python golden-file parity on Windows. Changes in version 0.99.1 - Initial Bioconductor submission candidate. - Provides primer inference, extraction, sparse multi-round counting, quality control, reproducibility manifests, FASTQ input, and ENA data discovery. - Adds selexprep_public_catalog as a documented, reproducibly generated package dataset with immutable snapshot provenance. - Adds R-native barcode demultiplexing and direct FASTQ/fetch-result input bridges with file hashes retained in the run manifest.