# Drug-response data splits: leave-cell-line-out, leave-drug-out, leave-tissue-out

Source: https://onco.cc/terms/drug-response-splits/  
OnCo record `drug-response-splits` (Term). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

How a drug-response dataset is split decides what a model's accuracy means: hold out cell lines to test personalised prediction, hold out drugs to test drug design, hold out tissues to test repurposing; holding out random pairs only tests imputation.

## Summary

Bernett and colleagues' DrEval framework evaluates drug-response prediction models under explicit protocols, holding out cell lines, drugs or tissues so that a model is scored on the generalisation it claims, and finds that many published gains shrink under the harder splits. Training on one screen and testing on another (cross-study validation) is the sternest bar because assays, cell-line panels and summary statistics differ. Random pair splits leak both the cell line and the drug into training and should only support imputation claims.

## Fields

- Kind: Term
- Last checked: 2026-09-24
- Also known as: leave-cell-line-out; LCO split; leave-drug-out; LDO split; leave-tissue-out; LTO split; leave-pair-out; random pair split; cross-study validation
- Tags: cansim-terms

## Notes

- Listed in the CanSim terms map 1.0.0 (docs/onco/terms.json, generated 2026-09-24), CC BY 4.0, attribution: CanSim project, an open, public-data-first cancer foundation-model programme; CanSim page path /terms/leave-cell-line-out.

## Sources

- Bernett et al., Critical evaluation of drug response prediction models with DrEval (Nature Communications 2026): https://doi.org/10.1038/s41467-026-72903-w

## Connected records

- terms: [Cancer AI vocabulary (CanSim terms map)](https://onco.cc/terms/cancer-ai-vocabulary/), [Data leakage in model evaluation](https://onco.cc/terms/data-leakage/), [Drug response and drug sensitivity (IC50, AUC)](https://onco.cc/terms/drug-response-sensitivity/), [Drug-response baselines and frameworks: mean-drug floor, LightGBM, DrEval, IMPROVE, DeepTTA](https://onco.cc/terms/drug-response-baselines/), [External validation](https://onco.cc/terms/external-validation/)

---
JSON: https://onco.cc/api/v1/entities/drug-response-splits.json