beiken1201's picture
Unify downstream task layout and document dataset launch commands
3830179 verified
|
Raw History Blame Contribute Delete
1.51 kB
# Drug-combination synergy data
This folder is a cleaned data layout for drug-combination synergy prediction.
The bundled legacy collaborator reference package is located at this repository-relative path:
```text
data/downstream/final_data/drug_response_pytorch/drug_combination
```
## Runnable data path
From the repository root, use this directory as `--data_dir` for `BioMod/drug_combination/train.py`:
```text
data/downstream/final_data/drug_combination_synergy/data_summary
```
Required files used by the current script:
```text
data_summary/
|-- new_labels_0_10.csv
|-- new_cell_features_954.csv
|-- smiles_0_10.csv
`-- test_data/
|-- independent_input.csv
`-- independent_cell_features_954.csv
```
## Previously recorded statistics
The figures below are retained from the existing dataset documentation; this path/comment audit did not recompute them.
- Training labels: (12415, 4)
- Training cell features: (32, 955)
- SMILES table: (35, 2)
- Independent test labels: (668, 4)
- Independent test cell features: (1020, 955)
- Training label counts: {'0': 6299, '1': 6116}
- Test label counts: {'1': 480, '0': 188}
- Unique training cells: 31
- Unique test cells: 24
- Unique training drug columns: drug1=35, drug2=35
- Unique test drug columns: drug1=44, drug2=39
Additional files in `data_summary/` are retained for provenance and regeneration of train/test labels.
## Files
See `MANIFEST.csv` for file sizes and checksums. See `SUMMARY.json` for machine-readable shape checks.