NCI60-2026 PharmacoSet2 recuration

Standalone Snakemake pipeline to build the NCI60-2026 PharmacoSet2 from the NCI April 2026 growth inhibition release, companion NCI Chemical Data files, and NCI Molecular Target Data plus selected CellMiner processed molecular profiles.

Setup

Install the pixi environment, then run the one-time GitHub package workaround:

pixi install
pixi run setup

The setup task installs bhklab/CoreGx@feat/sparse-longtable-index, bhklab/PharmacoGx@feat/mcp, and bhklab/AnnotationGx@main into the pixi R library. The CoreGx branch is used for sparse LongTable construction, and the PharmacoGx branch is used because the current devel branch resolves to an older slow curve fitter in this pixi R environment.

Source Files

Note: NCI currently blocks programmatic downloads from the configured wiki attachment URLs with 403 Forbidden. Download the source files manually from the NCI and CellMiner websites and place them under data/rawdata/growth_inhibition/, data/rawdata/chemical_data/, data/rawdata/molecular_profiles/, and data/rawdata/cellminer/ using the configured filenames. The pipeline still keeps automatic download logic for future working links; set download: true in config/pipeline.yaml to enable it.

Run

pixi run snakemake -c <cores>

The official config enables supplemental PharmacoGx profile recomputation from raw dose-response curves. For faster development runs, set recompute_profiles.enabled: false in config/pipeline.yaml.

Outputs:

• data/results/NCI60-2026_PSet.rds
• data/results/NCI60-2026_tables/

Raw source files are under data/rawdata/. Exact release details are stored in the PSet annotation metadata, and the manually inspected source-file inventory is documented in docs/data_sources.md.

Additional processing decisions are documented in docs/processing_decisions.md.

Review The Curated PSet

Render the user-facing QC report from an existing PSet:

pixi run qc

The report is written to qc/NCI60_PSet_QC.html. It summarizes the object structure, release metadata, sample and treatment coverage, sensitivity experiment counts, molecular profile assays, concentration-unit counts, recomputed profile comparisons when available, dose-count distributions, and example raw dose-response curves. This report is intended for users who want to inspect the curated object without downloading or parsing the raw NCI source files.

By default the report reads data/results/NCI60-2026_PSet.rds.