梯度升压
苦味
Boosting(机器学习)
范畴变量
随机森林
人工智能
决策树
机器学习
计算机科学
化学
品味
食品科学
作者
Can Chen,Haichao Deng,Huijie Wei,Yaqing Wang,Ning Xia,Jianwen Teng,Qisong Zhang,Li Juan Huang
出处
期刊:Research Square
日期:2024-05-30
标识
DOI:10.21203/rs.3.rs-4439031/v1
摘要
Abstract Establishing the bitterness threshold of molecules is vital for their application in healthy foods. Although numerous studies have utilized Mathematical algorithms to identify bitter chemicals, few models can accurately forecast the bitterness threshold. This study investigates the binding mode of bitter substances to the TAS2R14 receptor, establishing the relationship between the threshold and binding energy. Subsequently, a structure-taste relationship model was constructed using random forest (RF), extreme gradient boosting (XGBoost), categorical boosting (CatBoost), and gradient boosting decision tree (GBDT) algorithms. Results showed R-squared values of 0.906, 0.889, 0.936, and 0.877, respectively, suggesting a relatively good predictive capability for the bitterness threshold. Among these models, CatBoost performed optimally. The CatBoost model was then employed to predict the bitter thresholds of 223 compounds. The model provides a precise reference for detecting the bitterness thresholds of a wide range of chemicals and dangerous substances.
科研通智能强力驱动
Strongly Powered by AbleSci AI