Lab animal strains in published research: Methods sections, data version 2.0.0, built 2026-11-13 https://conductcolony.com/data/reporting How to cite ConductScience. (2026, November 13). Lab animal strains in published research: Methods sections [Data set]. https://conductcolony.com/data/reporting What is counted For each of the 131 strains, stocks, models and kinds of line in the strain list of the main data set (colony-data-1.0.0.json and colony-strains-1.0.0.csv, with the same ids and search terms): the papers whose Methods section names it, the papers whose title or abstract names it, and the papers where both do, for each year from 2001 to 2026. Every number in the file is a count of papers, never of animals. A paper counts once for an entry, however often it names it. A name counts even in passing, for example as the background strain of a knockout line or as the source of donor cells, so a name in the Methods does not prove that the paper used those animals. Which papers Every article in NLM's PMC Cloud Service on 2026-11-12: PubMed Central's open access subset and its author manuscripts, the newest version of each. NLM listed 9,338,054 articles that day. ScienceGraph read them once into its pmc-text release (pmc-text-2026-11-12-0.1.0), which holds 9,337,991 of them; 63 could not be read (for 51, NLM's record or file was missing when read, and 12 did not match NLM's checksum). This is not all of PubMed Central: articles that are neither under a license that allows reuse nor collected under a funder's policy, such as NIH's, are not in the set. Compare shares within this file, such as an entry's papers out of the papers with a Methods section in the same year, rather than its totals with other sources. Only research papers count: the article types research-article, brief-report and rapid-communication, and letters that have a Methods section. That gives 6,900,768 papers from 2001 to 2026, of which 1,008,246 are author manuscripts and 17,944 are marked retracted (they count, as they do in PubMed). Reviews, case reports, editorials, corrections and the other types are left out. The JSON file lists every type kept and left out, with counts for papers of any year. A paper's year is the year of its PubMed record. For the few papers with no PubMed record, it is the year of online publication, else of print publication, else of the issue. 2026 is a part year: its papers run to 2026-11-12, the day NLM listed the articles. Five papers give an online date after that day, a typing error or a date set ahead, and that date is never used. Methods sections A section of the paper's body or back matter that the publisher tags as methods or materials (never as supplementary material), or whose title is a Methods title such as Methods, Materials and methods, Experimental procedures, Animals, Online methods or STAR Methods, with all its subsections. A section inside Results, Discussion, Introduction, Conclusions or Background is never read as Methods. A paper has a Methods section when these sections hold at least 20 letters beyond their titles: 6,062,374 of the 6,900,768 papers do. Tables in the Methods are read. Formulas, images and supplementary files are not, so Methods kept only in a supplementary file are not seen. Email addresses in the text were replaced with "[email removed]" when the release was made; no search term holds an email address, so no match changes. How a match works The same search terms and rules as the title and abstract counts of the main data set. Names that have other meanings count only beside a word for the animal, and most entries also need their species to be found in the same text. In the Methods, a species is found when PubTator 3 (NCBI) labels it in words that appear in the Methods, or when the Methods use the species' own names, such as mice, murine or Mus musculus, because PubTator 3 has no full-text labels for some papers. The title and abstract are read from the paper's own file with the main data set's rule, so these counts can differ a little from the main data set's PubMed counts. A substrain counts on its own row and for its parent: a paper naming C57BL/6J also counts for C57BL/6. Denominators parsed: research papers in the year, all of them read. withMethods: those with a Methods section. mouseMethods: papers whose Methods name a mouse strain, stock or model. ratMethods: the same for rats. The kinds of line, such as knockout, transgenic or Cre-lox, count in any species and so are left out of these two. C57BL/6 Among papers whose Methods name C57BL/6 or one of its substrains: j names C57BL/6J (including forms such as C57BL/6JRj, C57BL/6JOlaHsd and C57BL/6JCrl), n names C57BL/6N (such as C57BL/6NCrl, C57BL/6NTac, C57BL/6NHsd, C57BL/6NJ and C57BL/6NRj), jAndN names both, and unstated names neither. Other substrains, such as C57BL/6ByJ, count as unstated. In every year, j + n - jAndN + unstated = named. RRIDs anyInMethods: papers whose Methods hold at least one Research Resource Identifier written after "RRID:", counting placeholders such as "RRID: N/A" too. strainRrid: papers whose Methods hold an RRID for an organism strain or stock, known by its prefix. Mice: IMSR from any repository (IMSR_JAX, IMSR_CRL, IMSR_TAC and the rest), MGI, MMRRC and EMMA, and a Jackson (JAX) or RIKEN (RBRC) stock number. Rats: RGD and RRRC. Zebrafish: ZFIN, ZIRC, and ZFIN's own ids for lines, fish, genotypes and transgenic constructs (ZDB-ALT, ZDB-FISH, ZDB-GENO, ZDB-TGCONSTRCT). Flies: BDSC, Bloomington, DGGR, DGRC, VDRC, FlyBase and its stock ids (FBst), Kyoto and NIG. Worms: WB-STRAIN (also written WBStrain) and CGC. Frogs: NXR, EXRC and XEP. Axolotls: AGSC. Pigs: NSRRC. Other stock centers: XGSC, TSC, NBRP and SRPC. Any prefix that starts with IMSR counts, however it is spelled after that (IMSRJAX, IMSR-JAX). Other misspelled prefixes, species ids (NCBITaxon), fly alleles and constructs, and cell line numbers do not count. strainRridAmongMouse: papers with a strain RRID whose Methods also name a mouse strain, stock or model. prefixes: the number of papers holding each prefix, strain or not (AB is an antibody, SCR software, CVCL a cell line, Addgene a plasmid). A stock number written without "RRID:" is not counted. Sex Among papers whose Methods name a mouse or rat strain, stock or model: male when the Methods use the word male or males and never female or females; female the other way round; both when they use both words, or a phrase such as both sexes, either sex, mixed-sex or sexes pooled; none when they use neither. Man, men, woman and women never count. A word counts anywhere in the Methods, so Methods that describe male and female breeders, or people as well as animals, count as both. Two rows to read with care SJL: most papers on this row name SJL only inside the name of a congenic line, such as B6.SJL-Ptprca, the CD45.1 mouse on a C57BL/6 background, rather than SJL mice. That holds for 3,829 of the row's 6,984 papers (1,455 of 2,415 from 2021 to 2025). CBA: 374 of the row's 2,983 papers (156 of 961 from 2021 to 2025) name only BD's cytometric bead array (CBA) kits. Reading check 100 Methods matches were drawn at random, at least one from each of the 50 entries with the most matches and more from the largest. Each was read in context and judged right (the Methods name the entry, or a member of it, as the animals used, as a source of cells, tissue or data, or as a background), wrong, or unsure. The JSON file's reading block gives the number of each verdict. Files colony-methods-2.0.0.json: the counts for each entry and year (methods; abstract, which counts the title or abstract; both), the denominators, the C57BL/6 split, RRIDs and sex by year, the papers read and left out, the reading check, and the SHA-256 of each file of code that made it. Years are text keys from "2001" to "2026", and every year is present, zeros included. In the corpus block, files and parsed count the articles read (one row each), not files, and researchOpenAccess and researchAuthorManuscripts overlap (an article can be in both sets), so together they come to more than researchArticles. The earlier version Data 1.0.0 (colony-methods-1.0.0.json and its readme, built 2026-11-12) counted NIH's copy of PubMed Central open access articles for the SPARK challenge, about 5% of PubMed Central, with the same rules. Its files stay where they were. The papers differ, so no count or share is comparable between the two versions. Sources NIH NLM NCBI PubMed Central (PMC) Article Datasets - Full-Text Biomedical and Life Sciences Journal Articles on AWS was accessed on 2026-11-13 from https://registry.opendata.aws/ncbi-pmc (the PMC Open Access Subset and the Author Manuscript Dataset, as NLM listed them on 2026-11-12). Each article keeps its own license or its funder's policy; this file holds counts and labels only, never article text. PubMed records (NLM); PubTator 3 (NCBI). Courtesy of the U.S. National Library of Medicine. NLM has not endorsed this data set or ConductColony.