Interpretation of machine learning models using shapley values: application to compound potency and multi-target activity predictions.
Rodríguez-Pérez R, Bajorath J. · 2020
Read Interpretation of machine learning models using shapley values: application to compound potency and multi-target activity predictions. by Rodríguez-Pérez R, Bajorath J. (2020) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc7449951.txt.
Description
Rodríguez-Pérez and Bajorath explain how to interpret machine learning models in drug research. Black-box models limit trust and use in pharmacy work. The SHAP method assigns each feature a share in a model's prediction.
SHAP works with any model, from tree ensembles to deep neural networks. It ranks the features that drive compound class and activity forecasts. The authors test an exact Shapley value method for decision trees. They compare it with the general model-free SHAP approach on potency tasks. New uses include reading deep network models that predict activity on many targets at once. Ensemble regression models for potency also get the SHAP treatment. The work aids clear, checkable model use.
Catalog record
| CREATOR | Rodríguez-Pérez R, Bajorath J. |
|---|---|
| YEAR | 2020 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 57 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |