Scholarlink

Accurate prediction of enzyme thermostability is crucial for bioengineering applications. This paper proposes a novel ensemble learning framework for predicting protein thermostability. The proposed ensemble learning framework combines XGBoost, a potent gradient boosting technique, with a Bidirectional Long Short-Term Memory (BiLSTM) network, which captures complex sequence-based features. The proposed framework attained the RMSE, MAE, R2 score, and Spearman Correlation coefficient of 0.37, 0.68, 0.72, and 0.76 respectively. Its performance is also evaluated against other machine-learning models and performs noticeably better than all of them. Furthermore, we leveraged Explainable Machine Learning (XML) techniques like SHAP (SHapley Additive Values), LIME (Local Interpretable Model Explainer), ELI5 and QLattice to enhance model interpretability.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enzyme Stability Prediction: Advancing with Ensemble Machine Learning and Explainable Artificial Intelligence

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Swati Sampada Parida

摘要

Accurate prediction of enzyme thermostability is crucial for bioengineering applications. This paper proposes a novel ensemble learning framework for predicting protein thermostability. The proposed ensemble learning framework combines XGBoost, a potent gradient boosting technique, with a Bidirectional Long Short-Term Memory (BiLSTM) network, which captures complex sequence-based features. The proposed framework attained the RMSE, MAE, R2 score, and Spearman Correlation coefficient of 0.37, 0.68, 0.72, and 0.76 respectively. Its performance is also evaluated against other machine-learning models and performs noticeably better than all of them. Furthermore, we leveraged Explainable Machine Learning (XML) techniques like SHAP (SHapley Additive Values), LIME (Local Interpretable Model Explainer), ELI5 and QLattice to enhance model interpretability.