QSAR and machine learning models of redox potentials of some organic pigments

Abstract

The organic pigments offer promising opportunities for developing new sustainable electrode materials for lithium batteries. Some of them have been identified as cathode material with very encouraging reversible lithium ion storage characteristics. One of them is a naturally occurring purpurin extracted from the Madder plant (Rubia tinctorum) for which we confirmed this good electrochemical behavior by cyclic voltammetry. One of the strategies towards obtaining materials with even better characteristics is a structural modification of already existing pigments. Building a theoretical model that could predict the redox properties of these new compounds can be very useful towards achieving that goal. In order to build a 3D QSAR (quantitative structure–activity relationship) model for material redox potential prediction, 9 organic pigments with known redox potentials were extracted from the literature. Based on molecular interaction field (MIF) probes we calculated standard GRIND (grid-independent) descriptors and constructed following principal PLS (partial least squares) model. By validation with the literature data, but also with the obtained experimental data for purpurin, this model proved very reliable in predicting the redox potential. A comparison was also made with the machine learning model that was formed in parallel.