DrugDis data

Processed inputs for the code at github.com/Boom5426/DrugDis, which accompanies DrugDis: Disentangling general and context-specific effects in drug-response prediction (project page). Download the whole dataset and point the code at it:

from huggingface_hub import snapshot_download
snapshot_download("Boom5426/DrugDis", repo_type="dataset", local_dir="/path/to/DrugDis-data")

export DRUGDIS_DATA=/path/to/DrugDis-data

Files

file	contents	built by (in the code repository)
droma.sqlite	the DROMA database of harmonized preclinical drug-response and omics data	DROMA (doi:10.5281/zenodo.18503188)
annotations/drug_anno_with_struc_info.csv, annotations/drug_anno_nci60_structure.csv	compound name to SMILES tables, as used to build drug_response.parquet	provided as used
master_table.parquet	one row per DROMA sample: project, model type, tumour type	drugdis/data/build_processed_tables.py
drug_response.parquet	one response per (compound name, sample): canonical SMILES, response value, resource	drugdis/data/build_processed_tables.py
gdsc_sensitivity_data.parquet	GDSC1 and GDSC2 responses kept apart; rows before 229,420 are GDSC1	drugdis/data/build_gdsc_sensitivity.py
omics_mrna_raw/	expression per cohort, samples × genes (.parquet, with .h5ad copies)	drugdis/data/build_processed_tables.py
omics_baseline/	the 15,961 genes shared by CCLE and GDSC, and all cohorts on those genes	drugdis/data/build_processed_tables.py
omics_baseline_frozen/	the CCLE-derived transcriptomic input matrix of the benchmark dataset (1,406 samples × 15,961 genes) and its provenance record	drugdis/data/build_ccle_substrate.py
Molecule_Embeddings/	ECFP4 (radius 2, 2,048 bits) and eleven pretrained molecular representations, as {canonical SMILES: vector} pickles	ECFP4: drugdis/data/encode_ecfp4.py; others: the published models cited in the manuscript
Gene_Embeddings/	measured expression (Baseline) and eleven pretrained transcriptomic embeddings	the published models cited in the manuscript

The benchmark dataset of the manuscript (3,141,680 drug–sample pairs, 986 cell lines, 54,180 compounds) is defined from these files by configs/substrate_config.frozen.json in the code repository, and the prespecified split manifests are in its manifests/.

Notes on construction

• Overlap rule. When the same compound name is measured on the same sample by more than one response resource, drug_response.parquet keeps the measurement of the first resource in a fixed order (CCLE, CTRP1, CTRP2, FIMM, GDSC1, GDSC2, GRAY, HKUPDO, LICOB, NCI60, PDTXBreast, Prism, Tavor, UHNBreast, UMPDO1, UMPDO2, UMPDO3, Xeva, gCSI).
• Tavor. DROMA types the 53 Tavor samples as PDC. master_table.parquet labels them "Cell Line", and the benchmark dataset excludes the Tavor project, so Tavor enters neither the cell-line benchmark nor the organoid cohorts.
• Organotin compounds. Two NCI60 organotin compounds carry a five-valent [Sn-] in their SMILES. Their fingerprints are stored under the neutralised [Sn] form, which matches no SMILES in drug_response.parquet, so neither compound is in the benchmark dataset. RDKit 2023.09.4 rejects the [Sn-] form, so a rebuild with it lacks their 109 NCI60 rows and their two fingerprints; every other row and fingerprint is reproduced exactly.
• The upstream resources are NCI60, PRISM, CTRP1, CTRP2, GDSC1, GDSC2, CCLE, GRAY, gCSI, FIMM and UHNBreast for cell lines, and UMPDO1, UMPDO2, UMPDO3, HKUPDO and LICOB for organoids. The input-data provenance of DROMA is doi:10.5281/zenodo.17498421.