Demonstrates machine learning accurately predicts olfactory thresholds in pyrazine derivatives, suggesting efficient flavor and fragrance development.
Key Points
The aim is to develop a machine learning model to predict olfactory thresholds of substituted 1,4-pyrazine derivatives based on their molecular structures.
Analyzed 78 pyrazine derivatives using cheminformatics descriptors and SMILES representations.
Evaluated multiple machine learning algorithms including Random Forest and Extra Trees for predictive accuracy.
Performed descriptor preprocessing and dimensionality reduction before modeling.
Extra Trees algorithm achieved the highest external accuracy with an R² value of 0.814.
Random Forest and Bagging followed with R² values of 0.802 and 0.784, respectively.
Models displayed strong generalization capability, indicated by low root mean square error (RMSE) values.