Advancement of predictive modeling of zeta potentials (ζ) in metal oxide nanoparticles with correlation intensity index (CII)

It was expected that index of the ideality of correlation (IIC) and correlation intensity index (CII) could be used as possible tools to improve the predictive power of the quantitative model for zeta potential of nanoparticles. In this paper, we test how the statistical quality of quantitative structure-activity models for zeta potentials (ζ, a common measurement that reflects surface charge and stability of nanomaterial) could be improved with the use of these two indexes. Our hypothesis was tested using the benchmark data set that consists of 87 measurements of zeta potentials in water. We used quasi-SMILES molecular representation to take into consideration the size of nanoparticles in water and calculated optimal descriptors and predictive models based on the Monte Carlo method. We observed that the models developed with utilization of CII are statistically more reliable than models obtained with the IIC. However, the described approach gives an improvement of the statistical quality of these models for the external validation sets to the detriment for the training sets. Nevertheless, this circumstance is rather an advantage than a disadvantage.