Quantitative models for predicting antioxidant capacity in herbs based on molecular structures and compositions

Abstract

Herbs are considered as a vital source of natural antioxidants that can neutralise free radicals which cause harmful health effects to the human body. Researchers have found that the phenolic compounds are the major phytochemicals in herbs that contribute to their antioxidant capacity. However, even though the herbs are grown in the same conditions and geographic origin, the components and composition of phenolic compounds may differ for each sample, contributing to different antioxidant capacities. Previous researchers have only studied the interactions between either their molecular structures or composition of phenolic compounds. The interaction and synergistic effect of the combined components and composition of phenolic compounds contributing to their antioxidant property are still unknown. The aim of this research is to understand the synergistic effect between the structure and composition of phenolic compounds in herbs by developing a quantitative model. Firstly, a Quantitative Structure-Activity Relationship (QSAR) model was developed in three different approaches, namely general, consensus and comprehensive models using literature data set of traditional Chinese medicine. Previous research have developed the QSAR models using all generated molecular descriptors without any classification that might overlooked the important variable. In this research, the general and consensus models were built using the molecular descriptors from the DRAGON software. The general model utilised all the molecular descriptors, while the consensus model classified the molecular descriptors according to the phenolic compound groups. In addition, quantum-chemical descriptors from the Gauss View 5.0 and Gaussian 09 software which were also added into the model to include 3D descriptors in the model, and therefore, the model is known as the comprehensive model. Then, a new Quantitative Structure-Composition-Activity Relationship (QSCAR) model was developed by using the experimental data set to further correlate between the molecular structure (from QSAR model) and composition ratio for each significant phenolic compound in Misai Kucing. Three variable selections, namely forward stepwise, interval-partial least square (i-PLS) and genetic algorithm and two multi-linear regression analysis methods were combined to developed all models. The best performance QSCAR model based on the robustness, reliability and predictivity was selected and the result was compared with QSAR model and experimental results. As a result, the consensus model produced overall performance better than the general model. The increment of antioxidant activity is correlated with the phenolic compound size through measurement of the bond indices distance between the atom, shape that is specifically calculated in the proportion of path/walk in length 3 from molecular Randic shape index and the number of bridge edges. The high ratio between EHOMO and ELUMO, the low of stability and total energy values of phenolic compounds increased the antioxidant activity as well. The QSCAR could predict the antioxidant capacity with 13.88 % more accurately than the QSAR model. The QSCAR model shows that the high compositions of apigenin and dalspinosin while the low composition of caffeic, ferulic and rosmarinic acids increased the antioxidant capacity in Misai Kucing. In conclusion, a quantitative model has been developed to predict the antioxidant capacity in herbs by combining the comprehensive QSAR and QSCAR models. The QSAR model is generic for phenolic compounds, but QSCAR needs to be simulated again with the other herb composition ratios. Thus, the future researchers can use the models to predict antioxidant capacity for other herbs. The research may also be beneficial by extending the model for predicting other biological activities.