krt

krt is a toolkit for authoring, validating, enriching, rendering, and depositing Key Resources Tables (KRTs): structured tables that list the resources a study used and generated (antibodies, cell lines, organisms, chemicals, software, datasets, protocols, and more), each paired with a persistent identifier so that every resource is unambiguously identifiable and machine-actionable.

KRTs began in Cell Press STAR Methods and are now required by funders such as ASAP (Aligning Science Across Parkinson’s). Once journals began requesting structured resource identifiers, reported analyses of RRID adoption found the fraction of antibodies a reader could unambiguously identify rose sharply, from roughly a quarter to nearly all.

krt is a standards orchestrator, not a template copier: it models resources around a neutral, typed core schema and maps them to journal or funder output profiles. The author-facing table is a view; the underlying record set stays structured, typed, and losslessly round-trippable through JSON and YAML.

Installation

# install.packages("pak")
pak::pak("choxos/krt")

Quick start

library(krt)

k <- new_krt("My study", study_type = "wet-lab")
k <- add_resource(k, "Antibody", "Rabbit Anti-TH",
                  vendor = "Millipore", catalog_number = "AB152",
                  rrid = "RRID:AB_390204", new_or_reuse = "reuse")
k <- add_resource(k, "Software/code", "Fiji", version = "2.14.0",
                  rrid = "RRID:SCR_002285", new_or_reuse = "reuse")

# Validate against a profile (ASAP is strict about identifiers)
validate_krt(k, profile = "asap")

# Normalize identifiers, then export to the ASAP six-column CSV
k <- normalize_ids(k)
export_krt(k, file.path(tempdir(), "resources.csv"), format = "asap")

# Or render a Markdown resource table for a manuscript
cat(render_krt(k, format = "md"))

What krt does

• Author typed resource records with a neutral core schema (14 ASAP resource types; identifiers stored by type, never conflated).
• Validate structurally and semantically, with conditional packs for cell lines (authentication and mycoplasma), organism metadata, software reproducibility, and human-subject ethics/consent. These are minimal, honestly scoped checks, not full ICLAC or ARRIVE assessments. Severity is tunable per profile.
• Normalize identifiers offline and optionally resolve them online (RRID, DOI, ORCID, PubMed, ROR, Cellosaurus), degrading gracefully when offline.
• Import from CSV, TSV, Excel, JSON, YAML, the ASAP six-column format, and Cell Press STAR three-column tables.
• Export and render to JSON/YAML (lossless), CSV/Excel/ASAP, RIS/BibTeX (with explicit lossy-export warnings), and Markdown/HTML/Word.
• Profiles: generic, asap (CC BY 4.0 assets), star-methods (interoperability), and your own; with programmatic license and attribution introspection (krt_audit_licenses()).
• Redact ethics and consent metadata for public-audience exports and all deposits (author exports are unredacted).
• Extract resources from manuscripts (PDF, JATS, DOCX, text) with a deterministic regex engine or an optional LLM.
• Deposit to Zenodo or Figshare, and connect to eLabFTW and protocols.io.
• Provenance: each editing step is recorded and exportable as PROV-JSON and RO-Crate 1.1.
• Interfaces: an R API, a command-line tool, a Shiny editor (launch_krt()), and an RStudio addin.
• Extensible through a plugin SDK (krt_plugin_api()) and Bioconductor-friendly via an S4 adapter.

Profiles and licensing

The package code is licensed GPL-3. The bundled ASAP profile assets are derived from the ASAP Key Resource Table resources (doi:10.5281/zenodo.17917979, CC BY 4.0) and are isolated under inst/extdata/profiles/asap/ with attribution and provenance. This package is independently developed and is not an official ASAP product; no endorsement by ASAP is implied. The STAR Methods profile contains only independently written rules; no Cell Press template is bundled. See LICENSE.note, inst/COPYRIGHTS, and krt_audit_licenses().

Use of AI

Parts of this package were developed with the assistance of AI coding tools. All code was reviewed and tested by the author. The optional LLM extraction engine sends manuscript text to a third-party provider only when the user explicitly runs it with their own API key; the deterministic regex engine is the default and runs entirely offline.

Citation

citation("krt")

License

GPL-3 for the package code; see LICENSE.note and inst/COPYRIGHTS for the licensing of bundled third-party materials.