Tracing NCI funding to shared data and its reuse by outside teams: analysis code and results

code/audit_paper_route_links.py

Seeded audit of the paper route's project-dataset links in the FY2026 publications-only group; writes the link accuracy result.

code/build_reuse_cohort.py

Harvests and merges the 2014 to 2025 reuse sample: week-stratified paper sample, accession annotations, production screen and repository check.

code/classify_reuse_outcomes.py

Extracts the passages naming each dataset in a sample of independent papers, applies keyword rules and recorded review decisions, and summarizes the uses.

code/correct_reuse_cohort.py

Applies cohort errata 1.0.1 (false datasets, co-producing papers, false matches, duplicate records) and 1.0.2 (rows the production review found cited rather than produced).

code/estimate_replacement_value.py

Estimates what collecting the independently reused data again would cost, from repository units and public price lists. Its default cache path is a folder on the authors' workstation; pass --cache.

code/measure_broad_reuse.py

Earlier reuse script; supplies the author-surname rule (surname_key) imported by the cohort and award-size scripts.

code/measure_dataset_discovery_routes.py

Earlier route-comparison script; supplies Europe PMC and NCBI helpers imported by the two other earlier scripts.

code/measure_draft_board_and_release_speed.py

Ranks datasets at age two against their later outside papers, finds late bloomers, and relates release timing to use.

code/measure_ins_coverage.py

Compares the Index of NCI Studies' dataset ties with the FY2026 portfolio groups and regroups the portfolio.

code/measure_paper_fame_and_clinical_reach.py

Tests whether producing-paper citations predict independent use, measures citation credit, and compares reusing papers' clinical citations with reference groups. Its default sample folder is on the authors' workstation; see README.

code/measure_payroll_vs_wins.py

Independent use and yield per NCI dollar by award size, investigator career stage, institution rank and cancer center status.

code/measure_portfolio_dollars.py

Places each FY2026 NCI core project in a group (official record, publications only, none) and sums the dollars; writes the award links table.

code/measure_reuse_beyond_nci_funding.py

Joins reusing US institutions to NCI grantees in NIH RePORTER and reads the funders acknowledged by each reusing paper.

code/measure_reuse_clock.py

Kaplan-Meier reuse clock, reuse by age, recent use and concentration, computed from the cohort file alone.

code/measure_reuse_per_dollar.py

Independent use per NCI dollar by grant type and institution type, from NIH RePORTER funding and publication counts.

code/measure_reuse_predictors.py

Signals visible at deposit and at year two that are associated with later independent use; model and held-out test.

code/measure_reuse_reach.py

Places independent papers by country, US state, institution type and topic using OpenAlex and the papers' own addresses.

code/measure_route_coverage.py

Earlier route-coverage script; supplies the accession allow-list (classify_accession) imported by the cohort scripts.

code/measure_share_score_predicts_nci_reuse.py

Tests whether SHARE 2.0 scores from the ShareScore public API predict independent use.

code/reuse_common.py

Shared helpers: API pacing and retries, key loading from environment variables, the core-project rule and complete Europe PMC listings.

code/review_unchecked_databank_production.py

Applies the repository production rule to the cohort's BioProject, SRA, dbGaP, PRIDE, EGA and ArrayExpress rows and recomputes the headline figures.

data/2026-09-24-nci-unseen-portfolio-dollars-v1.0.0.json

Result of measure_portfolio_dollars.py: FY2026 NCI dollars by group (official record, publications only, no traceable data). Key fields: export, definitions, headline, breakdown, byActivityGroup, routes, reconciliation, checks, perProjectTableSha256.

data/2026-09-26-ins-coverage-v1.0.0.json

Result of measure_ins_coverage.py: Index of NCI Studies dataset ties and the regrouped FY2026 portfolio. Key fields: inputs, howInsTiesDatasetsToProjects, insLinked, regrouped, unions, crossTab, publicationRouteAmongInsLinked, depositLevel, byActivityGroup, secondSourceSpotCheck, checks.

data/2026-09-26-paper-route-link-accuracy-v1.0.0.json

Result of audit_paper_route_links.py: accuracy of the paper route's links in the publications-only group. Key fields: population, sampleDesign, results, projectFollowUp, handVerification, dollars, links, checks.

data/2026-09-26-nci-reuse-cohort-v1.0.2.json

The reuse sample, cohort version 1.0.2 (schema 1.0.0): datasets produced by NCI-funded papers first published 2014 to 2025, as of 2026-09-23. Key fields: frame, sampling, rules, fieldNotes, funnel, datasets[] (accession, databank, originDate, originPaper, nciCoreProjects, originAuthorKeys, releaseCheck, outsidePapers, independentPapers), excludedByReleaseCheck, partLevelAccessions, removedByErrata, errata.

data/2026-09-26-unchecked-databank-production-review-v1.0.0.json

Result of review_unchecked_databank_production.py on cohort version 1.0.1; the basis of cohort errata 1.0.2. Key fields: rule, summary, calibration, handCheck, removalSets, errataCandidates, flaggedButKeptOnPageEvidence, headlinesBefore, headlinesAfterRecommendedErrata, downstream, rows.

data/2026-09-26-nci-reuse-clock-v1.1.0.json

Result of measure_reuse_clock.py: when outside use arrives. Key fields: pooled, byDatabank, byProductionYears, oldestCohort, papersPerDatasetYearByAge, newOutsidePaperInLast24Months, concentration, reviewerView, checks, bootstrap95.

data/2026-09-26-nci-reuse-early-signals-v1.1.0.json

Result of measure_reuse_predictors.py: early signals of independent use. Key fields: analysisSet, definitions, outcomes, univariate, adjusted, withinDatabank, model, validation, secondSource, datasets[].

data/2026-09-26-nci-reuse-reach-v1.1.0.json

Result of measure_reuse_reach.py (a recount on cohort 1.0.2): where independent papers come from. Key fields: method, headline, overall, byDatabank, countries, usStates, minorityServing, ideaStates, secondSource, recount, papers[].

data/2026-09-26-nci-reuse-outcomes-v1.1.0.json

Result of classify_reuse_outcomes.py: what independent teams did with the data. Key fields: selection, labels, labelGroups, summary, method, agreementCheck, rows[] (passages, ruleLabels, labels, primary, labelSource, review).

data/2026-09-26-nci-reuse-replacement-value-v1.1.0.json

Result of estimate_replacement_value.py: what collecting the reused data again would cost. Key fields: method, costTable, headline (conservative, milham), sensitivity, productionStatus, secondSource, datasets[].

data/2026-09-26-nci-reuse-per-dollar-v1.1.0.json

Result of measure_reuse_per_dollar.py: independent use per NCI dollar. Key fields: definitions, grants, tables.fractional (overall, byActivityGroup, byOrganizationType), tables.singleGrantPapers, concentration, secondSources, checks.

data/2026-09-26-payroll-vs-wins-v1.1.0.json

Result of measure_payroll_vs_wins.py: independent use and yield by award size, investigator career stage and institution. Key fields: base, definitions, institutions, main (tables, pointContrasts, contrastIntervals), singleGrantPapers, allFollowUp, checks.

data/2026-09-26-draft-board-and-release-speed-v1.1.0.json

Result of measure_draft_board_and_release_speed.py: ranking datasets at age two, late bloomers and release timing. Key fields: draftBoard (coverage, flaggedAtAgeTwo, heldOutAucByRanking), sleepers, releaseSpeed, checks.

data/2026-09-26-paper-fame-and-clinical-reach-v1.1.0.json

Result of measure_paper_fame_and_clinical_reach.py: producing-paper citations against later use, citation credit, and clinical citations of reusing papers. Key fields: part1 (rcr, citationCreditAllDatasets, citationCreditByUse, journalTierA, heldOut), part2 (reusePapers, comparisons), frameSample, pubmedBaselineSample, secondSource.

data/2026-09-26-reuse-beyond-nci-funding-v1.1.0.json

Result of measure_reuse_beyond_nci_funding.py: reusing institutions without NCI awards, and the funders reusing papers acknowledge. Key fields: method, part1 (headline, grantees, handCheck, matchAudit, institutions), part2 (headline), intervals95, checks.

data/2026-09-26-share-score-nci-reuse-v1.1.0.json

Result of measure_share_score_predicts_nci_reuse.py: SHARE 2.0 scores against independent use. Key fields: share, coverage, geo, signals, heldOut, auc, pooled, leakage, secondSource, datasets[].

requirements.txt

Python packages the scripts import, pinned to the versions used.