Big Data for a Deep Problem: Understanding and Predicting Antioxidant Interactions

Antioxidants are commonly added as combinations of two or more compounds, potentially improving their overall total antioxidant capacity. One drawback of this approach is that only a fraction (30%) of the reported antioxidant mixtures show synergistic effects, while the rest of the combinations in the literature feature either additive or antagonistic effects. An additional problem linked to the interaction between antioxidants is that the overall antioxidant capacity of these mixtures can be influenced by multiple and complex factors (distribution, matrix, concentration, ratio of antioxidants, pH, etc.), which often limits the prediction of the behavior. Approaches to understand these interactions have proposed the formation of antioxidant clusters, antioxidant regeneration due to redox cycling, differential localization of antioxidants in the sample, or a combination of different mechanisms (e.g., free radical scavengers and metal chelators).  While it is clear that no single mechanism can be used to predict the behavior of all antioxidant mixtures, several reports have pointed out the critical role of hydrogen bonding in improving antioxidant synergism and the difficulty to predict the consequences of this type of interactions. Aiming to address this gap in knowledge, our group has developed several machine learning algorithms that were trained to recognize hydrogen bonding and then fine-tuned to predict synergist interactions of antioxidants. Besides describing the evolution of these algorithms, the presentation will also discuss the possibility to form stable deep eutectic solvents (DES) with antioxidants and will highlight the importance of complementing predictions with experimental results. The applicability of the technology was demonstrated by determining the oxidative stability of oleic acid as well as commercial samples of olive oil, pork lard, and duck fat.
More information about this project can be found at: https://scienceweb.clemson.edu/uacl/