biobouncer 0.1.4

• Two new pattern sources for model-organism identifiers. mgi validates an MGI accession, such as MGI:97306, and rgd validates an RGD identifier, such as RGD:3059. Both require the prefix, since the bare number the registries record would accept any digits, and both repair a bare id or a lowercase prefix to the canonical form.

• Two new pattern sources for genomic variant coordinates. gnomad validates a gnomAD-style chrom-pos-ref-alt string, such as 1-55516888-G-A, with an optional chr prefix. spdi validates NCBI SPDI notation, such as NC_000001.11:55516887:G:A. Both check the coordinate shape offline in pattern mode, so an empty allele or a malformed field is rejected rather than passed through as valid.

• Documentation changes requested by CRAN on the first submission. The Description field now cites the Bioregistry and HGVS nomenclature references the identifier patterns follow, and the id_predicate() examples no longer carry commented-out code. The assertr, validate and pointblank integrations are still shown in the “Examples” article. There is no change to behavior.

biobouncer 0.1.3

• First CRAN submission. The maintainer ORCID and the copyright holder are now recorded in the package metadata. An internal helper calls utils::adist() through its namespace. There is no change to behavior.

biobouncer 0.1.2

• A lowercase gene symbol now suggests the right gene. hgnc suggestions in cache and existence mode ignore case, so tp53 suggests TP53 and brca1 suggests BRCA1. Case previously spent the edit budget, which left a symbol either with no suggestion (brca1 is four edits from BRCA1) or with the wrong one (tp53 is two edits from both TP53 and CD53, and the tie-break chose CD53). A lowercase typo resolves too: tp52 suggests TP53 rather than CD52.
• A case-wrong symbol stays invalid and carries the approved spelling as a suggestion, so report_id() and repair_id() clean it while an adapter still flags the cell. Suggestions use the snapshot’s own spelling, so C9ORF72 and c9orf72 both suggest C9orf72 rather than an uppercased form.

biobouncer 0.1.1

• No change to the R package. The R and Python packages share one version, and 0.1.1 is the first Python release published to PyPI.
• citation("biobouncer") now reports the installed version instead of a hard-coded 0.0.0.9000.

biobouncer 0.1.0

• Remote mode no longer writes an NCBI_API_KEY (or contact email) into the on-disk response cache or into error messages; the credential is redacted from the request URL before it is stored or shown.
• Cache mode and biobouncer_pull() reject a snapshot version that contains a path separator or .., so a version label cannot read or write outside the snapshot directory.
• New synthesize_ids() builds a synthetic, labeled column of identifiers for any source: a mix of valid, repairable, invalid, and missing values, each labeled by the checker in pattern mode (any source) or cache mode (a source that ships a snapshot, where a repairable value can be a retired id that maps to its successor). It is handy for exercising a validation pipeline (feed the column to report_id() or an adapter) without hand-writing test data, and it produces the same column as the Python synthesize().
• Rat (rattus_norvegicus) joins human and mouse as a covered species for the ensembl and uniprot sources, with pattern and remote corpus cases and recorded fixtures for the rat Tp53 gene and protein.
• Eight more sources now ship a bundled cache snapshot, so their cache mode works offline out of the box: bto, cl, doid, hp, mp, pato, so, and uberon. Each snapshot is a small set of real, current terms; a fuller snapshot is still available through biobouncer_pull().
• New opentargets source. It checks whether a human Ensembl gene id is a target the Open Targets Platform covers, through the platform’s GraphQL API. This is the first resolver that queries over a POST body rather than a URL.
• check_id() gains an on_error argument for remote and existence checks. With the default "raise" a network failure still unwinds the call; with "indeterminate" just that id is left NA with the reason in a new error column, and the rest of the batch is still checked. The result tibble now has that ninth error column, and report_id()/repair_id() accept on_error too.
• Checking a large column with remote mode can now run several requests at once. Set the BIOBOUNCER_REMOTE_WORKERS environment variable to the number of concurrent lookups (the default is 1, sequential). Concurrency only changes how fast the network is polled, never a verdict.
• New report_id() and repair_id() clean a whole column in one call. report_id() returns the check table classed so it prints with a one-line summary of how many values are valid, repairable, invalid, or missing; repair_id() substitutes the fixable values (a withdrawn gene symbol becomes its successor) and leaves valid, unmappable, and missing values untouched, so it drops into dplyr::mutate().
• cache mode no longer requires a version. When you omit it, biobouncer uses the latest installed snapshot (preferring a source’s pinned default), so a plain check_id(x, "hgnc", how = "cache") just works.
• existence mode now degrades to a pattern check for a source with no resolver, instead of raising, so it always returns a verdict.
• The adapters no longer treat a missing cell as a failure. check_valid_id(), assert_valid_id(), test_valid_id(), sv_biobouncer(), and id_predicate() now pass an NA value, matching the Python column checks, so a missing cell is never dropped by a filter or flagged by a data-frame rule.
• hgnc gene-symbol validation is now real in every mode. Cache mode ships a pinned approved-symbol snapshot of about 45,000 symbols, so it validates real gene symbols offline out of the box; a withdrawn or previous symbol still maps to its approved successor.
• biobouncer_pull("hgnc") now works. It refreshes the snapshot from a dated HGNC “complete set” archive through a new non-OBO builder, and the version label is the archive’s release date.
• Remote mode now covers hgnc. It checks live existence against the genenames.org service, so source_info() reports pattern, cache, and remote for it.
• Cache mode suggests the nearest approved symbol for a typo. A well-formed symbol that is neither approved nor a known previous symbol falls back to the closest approved symbol within a small edit distance, so TP52 suggests TP53.
• check_id() and is_valid_id() gain a refresh argument for remote and existence checks. When TRUE, a cached response is ignored and the id is looked up live again.
• Remote checks now record when each response was retrieved. A cached verdict reports its original fetch time in the version column instead of the time of the current run. Set the BIOBOUNCER_REMOTE_TTL environment variable to a number of seconds to refetch a cached response once it is older than that.
• Snapshot and remote-cache files are written atomically, so an interrupted write can no longer leave a truncated file that reports valid ids as invalid.
• Remote checks retry a transient network failure or a 429 or 5xx response a few times with exponential backoff before giving up, so a brief blip no longer fails a whole batch.
• The NCBI E-utilities checks (refseq, clinvar) send an API key when the NCBI_API_KEY environment variable is set, which raises NCBI’s rate limit from three to ten requests a second. NCBI_EMAIL is included when set.
• ncbitaxon is a new source for NCBI Taxonomy identifiers such as NCBITaxon:9606. It checks the pattern offline and existence against NCBI Taxonomy in the Ontology Lookup Service, reusing the OLS resolver. A bare taxon number or a different-case prefix is suggested in canonical CURIE form. No cache builder is offered, since the NCBI Taxonomy OBO release is far too large to snapshot.
• inchikey is a new source for InChIKey chemical structure keys such as BSYNRYMUTXBXSQ-UHFFFAOYSA-N. This is pattern mode only; a lowercase input is suggested in its canonical uppercase form. An existence check would need a UniChem or PubChem lookup, which is not offered yet.
• ncit is a new source for NCI Thesaurus concept codes such as NCIT:C3224, and eco for Evidence and Conclusion Ontology terms such as ECO:0000269. Both check the pattern offline and existence against their ontology in the Ontology Lookup Service, reusing the OLS resolver.
• Five InterPro member databases are new sources: smart (SM00248), panther (PTHR11003), cdd (cd00029), prints (PR00001), and ncbifam (TIGR00001 or NF000001). Each checks the pattern offline and existence against the EBI InterPro API, reusing the interpro resolver.
• mirbase_hairpin is a new source for miRBase hairpin precursor accessions such as MI0000001, a sibling to the mature mirbase source. It checks the pattern offline and existence against RNAcentral through EBI Search, reusing the mirbase resolver.
• mirbase gains remote mode. A mature accession is checked for existence against RNAcentral through EBI Search, which indexes miRBase. miRBase has no existence API of its own. This is an existence check only; a withdrawn accession is reported as absent, not with a successor.
• prosite gains remote mode. A pattern or profile accession is checked for existence against the ExPASy PROSITE entry endpoint, which resolves both entry types from one address. This is an existence check only; a deleted accession is reported as absent, not with a successor.
• orphanet gains remote mode. A rare-disease id is checked against the Orphanet Rare Disease Ontology in the Ontology Lookup Service, reusing the OLS resolver. The ORPHA prefix is rewritten to the ontology’s Orphanet prefix for the lookup, and an obsolete term is reported with its successor.
• clinvar gains remote mode. An accession is checked for existence by searching ClinVar through NCBI E-utilities. One search covers all three accession types (VCV, RCV, and SCV). This is an existence check only.
• refseq gains remote mode. An accession is checked for existence against NCBI E-utilities, routed to the nucleotide or protein database by its molecule prefix. The summary endpoint returns an empty result for an unknown accession. This is an existence check only; a suppressed accession is not distinguished from a current one.
• rfam, uniparc, complexportal, and wikipathways gain remote mode. Each checks existence against its source: the Rfam API, the UniProt UniParc endpoint, the EBI Complex Portal web service, and the published WikiPathways asset. All four are existence checks; an absent id is reported as not valid with no successor.
• interpro and pfam gain remote mode. An accession is checked for existence against the EBI InterPro API, which hosts both databases, so one resolver serves the two sources. The entry endpoint answers 204 for a well-formed accession that is not a current entry. This is an existence check only; a deleted accession is reported as absent, not with a successor.
• chembl gains remote mode. A ChEMBL id is checked for existence against the ChEMBL id-lookup endpoint, which resolves an id of any entity type, so one lookup covers compounds, targets, assays, and documents alike. This is an existence check only; an obsolete id is not yet reported with a successor.
• reactome gains remote mode. A stable id is checked for existence against the Reactome content service. This is an existence check only; a superseded stable id is not yet reported with its successor.
• pdb gains remote mode. A four-character structure id is checked for existence against the RCSB PDB data API. This is an existence check only; an obsoleted structure that was superseded is not yet reported with its successor.
• The Python package installs a biobouncer command-line tool. It validates identifiers from arguments, a file, or standard input, prints per-id results as text, TSV, or JSON, and exits non-zero when any input is invalid, so it drops into shell pipelines and CI. It also has biobouncer sources and biobouncer info.
• The Python package adds a Great Expectations column-map expectation, biobouncer.gx.ExpectColumnValuesToBeValidId, for validating a data frame column. Install it with pip install "biobouncer[gx]".
• id_predicate() now documents its use as a pointblank col_vals_expr() step, alongside assertr and validate. The predicate is unchanged; pointblank consumes it directly.
• First numbered release. Sets the version to 0.1.0 in both the R and Python packages so they track together.
• source_info() gains an example identifier and a modes column for each source, so it now answers “what does a valid id look like and how can I check it?”. The Python package gains a matching source_info().
• Adds a documentation website. The R reference and the getting-started vignette are built with pkgdown, the Python guide and API reference with MkDocs, and a shared landing page links the two. A docs workflow builds both on every pull request and publishes them to GitHub Pages from main.
• Adds the OBO ontology sources pato, mp, and bto for qualities (PATO:0000001), mammalian phenotypes (MP:0001262), and tissues (BTO:0000759), each with pattern and live OLS remote modes.
• Adds the OBO ontology sources so, hp, doid, uberon, and cl for sequence features (SO:0000704), phenotypes (HP:0001250), diseases (DOID:9352), anatomy (UBERON:0002107), and cell types (CL:0000236). Each supports pattern mode with prefix and zero-pad suggestions and live remote existence through the Ontology Lookup Service.
• Adds uniparc, complexportal, cosmic, and pharmgkb pattern sources for unique protein sequences (UPI0000000001), macromolecular complexes (CPX-2158), somatic mutations (COSM476), and PharmGKB accessions (PA267). Each suggests the canonical uppercase for a lowercase input.
• Adds wikipathways, orphanet, and mirbase pattern sources for pathways (WP554), rare diseases (ORPHA:558), and mature microRNAs (MIMAT0000001). Each suggests the canonical uppercase for a lowercase input.
• Adds drugbank, clinvar, and ec pattern sources for drug accessions (DB00001), ClinVar records (VCV000012345), and Enzyme Commission numbers (1.1.1.1). The first two suggest the canonical uppercase for a lowercase input.
• Adds pfam, rfam, and prosite pattern sources for protein families (PF00001), RNA families (RF00001), and PROSITE patterns and profiles (PS00001). Each suggests the canonical uppercase for a lowercase input.
• Adds reactome and interpro pattern sources. reactome checks Reactome stable ids such as R-HSA-68886, and interpro checks InterPro ids such as IPR000001. Both suggest the canonical uppercase for a lowercase input.
• Adds two more pattern sources. chembl checks ChEMBL identifiers such as CHEMBL25, and pdb checks Protein Data Bank structure ids such as 4HHB. Both suggest the canonical uppercase for a lowercase input.
• dbsnp gains remote mode. An rsID is checked for existence against the NCBI dbSNP RefSNP API. An rsID that was merged into another still resolves but is not current, so it is not valid and suggests the primary id.
• Adds two pattern sources. refseq checks NCBI RefSeq accessions such as NM_000546.6, with an optional version, and suggests the canonical uppercase for a lowercase input. dbsnp checks dbSNP reference SNP ids such as rs7412, and suggests lowercase for an uppercase prefix.
• hgvs gains remote mode. A variant is checked against the Mutalyzer normalizer, which confirms the reference sequence exists and the change is consistent with it, such as the stated reference base and coordinates in range. This goes beyond the offline syntax check. Only inputs that pass the offline grammar are looked up, and the response is cached on disk.
• Adds an hgvs source that checks the syntax of HGVS sequence variant names, for example NM_004006.2:c.4375C>T or NP_003997.1:p.(Gly56Ala). This is a grammar check in pattern mode: it covers substitutions, deletions, duplications, insertions, deletion-insertions, inversions, and the common protein forms including frameshifts, across the g, o, m, c, n, r, and p coordinate types. It checks syntax only. It does not check coordinate order or that a variant exists, and it requires a reference sequence.
• Adds validation-framework adapters that wrap the core classifier: assert_valid_id(), check_valid_id(), and test_valid_id() in the checkmate style, sv_biobouncer(), a shinyvalidate rule, and id_predicate(), an elementwise predicate for data-frame validation with assertr or validate. The Python package pairs these with a pandera check.
• Adds an hgnc source for HUGO gene symbols. cache mode checks a symbol against the approved-symbol snapshot, and a withdrawn or previous symbol resolves to its approved successor through the retired-map (for example MLL suggests KMT2A). Symbols are case-sensitive.
• Retired identifiers are detected with a successor suggestion. In remote mode an OLS term that exists but is obsolete is invalid and suggests its replaced_by successor. In cache mode a snapshot can carry a <version>.retired.tsv sidecar; an id retired in that version is invalid and suggests its successor. A cross-ontology successor is suggested as-is.
• remote mode is species-aware. An id that exists but belongs to a different species than requested is invalid: Ensembl is checked from its id prefix and UniProt from the entry’s organism taxon id.
• pattern mode is species-aware for Ensembl. When species is given, a well-formed id whose encoded species does not match is invalid, and a malformed id is only suggested when the correction matches the species. A species outside the source map is not checked.
• check_id() gains existence mode: it answers from a pinned snapshot when one is available for the requested version and otherwise falls back to remote.
• check_id() gains live remote mode: existence checks against a source API. The Ontology Lookup Service resolver covers mondo, efo, go, and chebi. Responses are cached on disk, and a network failure raises an error rather than returning a silent FALSE.
• remote mode adds Ensembl and UniProt resolvers, so ensembl and uniprot ids can be checked live. A retired (deleted) UniProt accession is reported as not valid.
• check_id() gains offline cache mode: existence checks against a pinned snapshot. A small sample snapshot for the ontology sources ships with the package, and biobouncer_cache_dir() and biobouncer_snapshots() manage snapshots.
• biobouncer_pull() downloads a full snapshot from a source’s OBO release into the cache directory (mondo, efo, go, chebi).
• check_id() and is_valid_id() implement offline pattern mode for an initial set of sources (mondo, efo, go, chebi, ensembl, uniprot).
• sources() and source_info() list what can be checked.
• Arguments are validated with checkmate, errors are raised with cli and carry a condition class, and check_id() returns a tibble.
• Initial package scaffold.