# Drug-response baselines and frameworks: mean-drug floor, LightGBM, DrEval, IMPROVE, DeepTTA

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

## TL;DR

The floor for any drug-response model is predicting each drug's average effect across cell lines; tuned gradient-boosted trees such as LightGBM often tie or beat deep models under honest splits, which is why evaluation frameworks now exist.

## Summary

LightGBM is Microsoft's free gradient-boosting framework based on decision trees (Wikipedia) and a common strong baseline. DrEval (Bernett and colleagues) and the IMPROVE project from Argonne and the NCI (with models such as UNO and GraphDRP) benchmark deep drug-response models against naive baselines under fixed protocols; the naive mean-drug predictor is hard to beat because most variance in a screen is between drugs, not between cell lines. Transformer models such as DeepTTA (Jiang and colleagues) report headline correlations that fall under leave-cell-line-out and cross-study splits.

## Fields

- Kind: Term
- Last checked: 2026-09-24
- Also known as: naive mean-drug baseline; mean-cell-line baseline; DrEval; IMPROVE benchmark; GraphDRP; UNO model; DeepTTA; DeepCCDS; LightGBM; gradient-boosted trees
- 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/drug-response-frontier.

## Sources

- Wikipedia: https://en.wikipedia.org/wiki/LightGBM
- Bernett et al., DrEval (Nature Communications 2026): https://doi.org/10.1038/s41467-026-72903-w
- IMPROVE project (Argonne and NCI) on GitHub: https://github.com/JDACS4C-IMPROVE
- Jiang et al., DeepTTA: a transformer-based model for predicting cancer drug response (Briefings in Bioinformatics 2022): https://doi.org/10.1093/bib/bbac100
- Wikipedia: https://en.wikipedia.org/wiki/LightGBM

## Connected records

- terms: [Benchmarks, leaderboards and contamination](https://onco.cc/terms/leaderboard-benchmark/), [Cancer AI vocabulary (CanSim terms map)](https://onco.cc/terms/cancer-ai-vocabulary/), [Drug-response data splits: leave-cell-line-out, leave-drug-out, leave-tissue-out](https://onco.cc/terms/drug-response-splits/), [Ridge regression and the multilayer perceptron](https://onco.cc/terms/ridge-regression/), [Spearman rank correlation](https://onco.cc/terms/spearman-correlation/)

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