Interpretable Machine Learning in Enzyme Classification: A SHAP-Guided Analysis of Global Structural Features
Journal of Artificial Intelligence Practice · 2026 · 9 (1)
DOI: 10.23977/jaip.2026.090113 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Artificial Intelligence Practice, 2026.
Abstract (Originalsprache)
Accurate computational annotation of protein function addresses a critical bottleneck in bioinformatics. This study presents an explainable machine learning framework that predicts a protein's functional class from its physicochemical profile. A Random Forest model, trained on four structural descriptors from the Protein Data Bank, classifies proteins into three major enzyme classes: Hydrolase, Oxidoreductase, and Transferase. The model achieved robust performance (test accuracy: 71.42%), with SHAP analysis identifying molecular weight as the primary discriminative feature. To ensure practical utility and reproducibility, the model is deployed as an interactive Shiny application and an open-source R package, providing a reliable and accessible tool for the community.
Zitieren
APA
(2026). Interpretable Machine Learning in Enzyme Classification: A SHAP-Guided Analysis of Global Structural Features. Journal of Artificial Intelligence Practice, 9(1). https://doi.org/10.23977/jaip.2026.090113
Vancouver
Interpretable Machine Learning in Enzyme Classification: A SHAP-Guided Analysis of Global Structural Features. Journal of Artificial Intelligence Practice. 2026;9(1). doi:10.23977/jaip.2026.090113
BibTeX
@article{euvimed2026,
title = {Interpretable Machine Learning in Enzyme Classification: A SHAP-Guided Analysis of Global Structural Features},
journal = {Journal of Artificial Intelligence Practice},
year = {2026},
volume = {9},
number = {1},
doi = {10.23977/jaip.2026.090113}
}
EUVIMED