CTRPv2 PharmacoSet

Overview

This repository curates CTRPv2 into an integrated PharmacoSet. CTRPv2.0 supplies the canonical raw dose-response experiments, sample metadata, and compound metadata. CTRPv2.1 supplies published AUC values, Affymetrix-derived microarray expression, and copy-number profiles.

Quick Start

pixi install
pixi run setup
pixi run pipeline
pixi run qc
pixi run zipup

Expected first-run runtime is 6-12 hours after package setup, depending on available cores and live AnnotationDB and AnnotationGx response times. Snakemake reuses declared outputs on incremental runs.

Pipeline

The staged workflow downloads source files, builds normalized dose-response chunks, enriches sample and treatment metadata, resolves current Ensembl gene IDs, recomputes PharmacoGx outcomes, constructs the PharmacoSet, and exports TSV alternatives for non-R users. Rules are organized by stage under workflow/rules/, with processing logic under workflow/scripts/.

Inputs

All source URLs and release settings are defined in config/pipeline.yaml. The pipeline downloads:

• CTRPv2.0_2015_ctd2_ExpandedDataset.zip
• CTRPv2.1_2016_pub_NatChemBiol_12_109.zip
• GENCODE v50 annotation and Entrez metadata
• Human NCBI Gene history and gene2ensembl records

Source provenance and the integrated-release scope are documented in docs/data_sources.md and docs/devnotes.md.

Outputs

• data/results/CTRPv2_PSet.rds: canonical PharmacoSet.
• data/results/CTRPv2_tables/: sample, treatment, gene, sensitivity, recomputed-profile, molecular-feature, and CTRPv2.1 AUC TSVs.
• data/results/CTRPv2_tables.tar.gz: compressed TSV bundle created by pixi run zipup.
• data/procdata/metadata/: enriched metadata and the GENCODE feature map.
• data/procdata/annotation_audit/: annotation misses and molecular features grouped by assay and exclusion reason.
• data/procdata/recomputed_profiles/: dose-response and fit/QC chunks.
• docs/CTRPv2_qc.html: rendered QC report.

QC and Docs

QC compares published CTRP values with PharmacoGx fitted-curve AUC and AAC, relative IC50, and absolute AC50. It also reports annotation coverage, fit quality, current-Ensembl rowname validity, molecular missingness, and excluded features. The public gene annotation TSV retains all source genes for reference, while molecular assays contain features with unique current Ensembl rownames.

The pipeline uses --cores all. The configured recompute chunk count controls scheduling granularity, available Snakemake cores control simultaneous fit workers, and annotationgx.workers controls AnnotationGx cell-line queries.

Development

pixi run setup installs the CoreGx, PharmacoGx, and AnnotationGx development refs required by the curation. Pipeline behavior is defined by versioned config and explicit code defaults. Format Python with uvx ruff format and R with air format after edits.