app/lib/publications/accessions.mjs

Functions

Imports

It imports no other module of this repo. Imported by app/lib/publications/validate.mjs, app/routes/publications.$slug.tsx.

Source

1// Accessions are read only inside the data-availability statement: a bare regex over a PDF returns
2// comparison phages and other outbreaks' deposits, a wrong citation. The text is the committed
3// extraction, which check:machine-readable binds to each PDF's sha256.
4
5// Case-insensitive with an optional statement: covers both ASM templates, Frontiers and MDPI.
6const SECTION_START = /\bdata availability\b\.?\s*(statement)?\s*:?\s*/i;
7
8// Cut at the next section so the sweep never reaches the reference list; MAX_SECTION bounds a
9// statement with no terminator.
10const SECTION_END =
11  /\b(acknowledg|references|supplemental material|author contributions|conflict of interest|funding|ethics|figures?\s+s?\d)/i;
12
13const MAX_SECTION = 900;
14
15// Per registry because the URL differs: one builder would link an SRA run to nuccore, a 404. The version
16// suffix is dropped: NCBI resolves to the current one and the statement does not pin a revision.
17const GRAMMARS = [
18  { kind: "bioproject", re: /\bPRJ[A-Z]{2}\d{4,}\b/g, url: "https://www.ncbi.nlm.nih.gov/bioproject/" },
19  { kind: "biosample", re: /\bSAM[A-Z]\d{4,}\b/g, url: "https://www.ncbi.nlm.nih.gov/biosample/" },
20  { kind: "sra", re: /\bSR[XRPS]\d{5,}\b/g, url: "https://www.ncbi.nlm.nih.gov/sra/" },
21  {
22    kind: "genbank",
23    re: /\b(?:NC_\d{6}|[A-Z]\d{5}|[A-Z]{2}\d{6}|[A-Z]{2}\d{8})(?:\.\d+)?\b/g,
24    url: "https://www.ncbi.nlm.nih.gov/nuccore/",
25  },
26];
27
28/**
29 * @typedef {object} Accession
30 * @property {string} kind one of the GRAMMARS kinds
31 * @property {string} id unversioned
32 */
33
34/**
35 * @param {string[]} pages the extracted text, one string per page
36 * @returns {Accession[]} sorted, so the output is a property of the paper
37 */
38export function accessionsInText(pages) {
39  const flat = (Array.isArray(pages) ? pages : []).join(" ").replace(/\s+/g, " ");
40  const start = SECTION_START.exec(flat);
41  if (!start) return [];
42
43  const after = flat.slice(start.index + start[0].length, start.index + start[0].length + MAX_SECTION);
44  const end = SECTION_END.exec(after);
45  const section = end ? after.slice(0, end.index) : after;
46
47  /** @type {Map<string, Accession>} */
48  const found = new Map();
49  for (const { kind, re } of GRAMMARS) {
50    for (const match of section.matchAll(re)) {
51      const id = match[0].replace(/\.\d+$/, "");
52      // First grammar wins; longest first, so an SRA run is claimed before the GenBank shape sees it.
53      if (!found.has(id)) found.set(id, { kind, id });
54    }
55  }
56
57  return [...found.values()].sort((a, b) => (a.id < b.id ? -1 : a.id > b.id ? 1 : 0));
58}
59
60/**
61 * @param {Accession} accession
62 * @returns {string}
63 */
64export function accessionUrl(accession) {
65  const grammar = GRAMMARS.find((g) => g.kind === accession.kind);
66  if (!grammar) {
67    // Never guess a registry: the wrong database 404s or, worse, shows a different record.
68    throw new Error(`no registry URL for accession kind ${JSON.stringify(accession.kind)}`);
69  }
70  return `${grammar.url}${accession.id}`;
71}
72
73const KIND_LABEL = {
74  genbank: "GenBank",
75  sra: "SRA",
76  bioproject: "BioProject",
77  biosample: "BioSample",
78};
79
80/**
81 * @param {string} kind
82 * @returns {string}
83 */
84export function accessionLabel(kind) {
85  const label = KIND_LABEL[/** @type {keyof typeof KIND_LABEL} */ (kind)];
86  if (!label) {
87    throw new Error(
88      `no label for accession kind ${JSON.stringify(kind)}. Add one here, or ` +
89        `the page and the registry would disagree about what it is.`,
90    );
91  }
92  return label;
93}