Elucidating Quantum Semi-empirical Based QSAR, for Predicting Tannins' Anti-oxidant Activity with the Help of Artificial Neural Network

Tannins are potential curatives, besides being an effective antioxidants. Here, tannin based QSAR with machine learning pipeline is elucidated. IC50 values of tannins' antioxidant activity were adapted from literature. This was further split into training and testing datasets. Furthermore, quantum semi-empirical descriptors were computed. Out of 277 chemical descriptors, 17 were shortlisted by feature selection Multiple Linear Regression. For the test dataset; R2 = 0.706 and mean absolute error (MAE) = 1.94. For the same dataset using nonlinear artificial neural network (ANN), R2 = 0.858 and MAE = 1.02. Therefore, AMPAC-CODESSA's feature selection and ANN, provides an efficacious tannin-QSAR model aiding tannin-based therapeutic design in future.

Automated summary

This paper presents a tannin-based QSAR and machine learning pipeline, which uses quantum semi-empirical descriptors and feature selection with a nonlinear artificial neural network. The model achieved good performance in predicting the antioxidant activity of tannins, providing guidance for tannin-based therapeutic design in the future.