Skip to content

CLAUDE.md

Latest commit

 

History

History
71 lines (55 loc) · 4.68 KB

CLAUDE.md

File metadata and controls

71 lines (55 loc) · 4.68 KB

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Purpose

CABBAGE (Comprehensive Assessment of Bacterial-Based AMR prediction from GEnotypes) collects, extracts, and curates all publicly available data pairing bacterial genotypes with AMR phenotypes (R/I/S or MIC values). It covers both major databases (CDC, COMPARE-ML, NARMS, NDARO, PathogenWatch, PATRIC, PubMLST) and individual papers from a literature search restricted to the WHO Priority Pathogen List (2017).

Running the Code

All scripts are R, intended to be run interactively from RStudio (project file: AMRDataTables.Rproj) or sourced into an R session. There are no automated build, lint, or test commands. Scripts are semi-manual: they define functions at the top, and the bottom of each file contains calls that require manual data entry (URLs, file paths) before executing.

Load order — source these files into a single R session in order:

  1. Utilities.R — shared functions and constants; must be loaded first
  2. ExtractInfo.R — genotype extraction functions
  3. PrepareTables.R — merging raw tables from papers
  4. PreprocessAllPapers.R — first-wave literature search processing
  5. PreprocessNewPapers.R — second-wave literature search processing
  6. PreprocessDatabases.R — database-specific processing
  7. ParseTables.R — AMR package antibiotic table utilities

Key Global Constants (defined in Utilities.R)

  • WORKING_DIR — absolute base path; currently hardcoded to ~/Downloads/AMR/AMRDataTables/. All file I/O is relative to this.
  • GOOD_SPECIES / GOOD_GENERA — the target organism list (WHO Priority Pathogens). Any filtering by organism traces back to these vectors.
  • ACCESSIONS — named regex vector mapping accession type names (RUN, ASSEMBLY, BIOSAMPLE, etc.) to their ENA/NCBI patterns.
  • FIELDS — defines the canonical schema: GENOTYPE (Database, Accession, Type), PHENOTYPE (Status, Antimicrobial, Class, MICValue, …), METADATA (Species, PubMedID, …).

Data Pipeline Architecture

Raw source files live in per-database or per-paper subdirectories. The pipeline produces files with standardized suffixes:

Suffix Meaning
_Processed.csv Raw data cleaned and reshaped into a standard wide format
_Merged_Processed.csv Multiple processed files concatenated
_SplitRes_Processed.csv Resistance profile column expanded into per-drug columns
_WidenedRes_Processed.csv Long-format resistance pivoted to wide (one column per drug)
_Mapped.csv Accession IDs resolved to run-level ENA accessions
_Extracted_Genotypes.csv Genotype accessions extracted; written to AllExtractedGenotypes/

Core functions in Utilities.R:

  • findAccessions() — scans all character columns of a table for ENA/NCBI accession patterns; returns the highest-priority match per row.
  • getGenotypes() — wraps findAccessions, adds Database column, writes *_Extracted_Genotypes.csv.
  • getAllGenotypes(Dir) — batch-runs getGenotypes over all *_Processed.csv files in a directory.
  • joinGenoPheno() — joins a genotype file and a phenotype file on a shared ID column.
  • concatenateFiles() — row-binds multiple processed CSVs with type coercion.
  • widenResistance() / splitResistance() — reshape resistance columns.
  • convertAccessions() — calls EBI API to resolve sample-level accessions to run-level.
  • downloadStudyDetails() — fetches ENA filereport for a study/sample accession.

File Naming Conventions

  • Papers: PMID_<id>_<supplement_filename>_<role>_Processed.csv (role = Genotype or Phenotype)
  • Preprints: DOI_<doi_with_colons_as_is>_…_Processed.csv
  • Databases: <DatabaseName>_<species_or_tag>_…_Processed.csv
  • PathogenWatch collections: pathogenwatch-<species>-<collection-id>-<author-year>_…
  • PubMLST collections: <Species>_BIGSdb_<collection-id>_…

Directory Layout

  • SelectedTables/ — processed tables for the first-wave literature search
  • RemainingTables/ — processed tables for lower-priority papers
  • AllExtractedGenotypes/ — one *_Extracted_Genotypes.csv per source, result of getAllGenotypes
  • Mappings/ — EBI-resolved accession mapping files (*_Mapped.csv) and consolidated AllMappings*.csv
  • IntermediateFiles/ — scratch files produced mid-pipeline
  • CABBAGEdata/ — final curated outputs including FinalDatasets.csv, FoundPapers.csv, SelectedPapers.csv
  • NDARO_Unprocessed/ — raw NDARO download files before processing
  • PATRIC/ — raw PATRIC genome metadata and AMR files
  • CARD Database/ — CARD reference data
  • NewTables/ — supplementary files for new/additional papers