QSPR MODELS FOR ZETA-POTENTIAL OF NANO-OXIDES PREDICTION

Nano-QSPR modeling often requires considering variety of factors, if neglected, may lead to erroneous result the study. Frequently, data turned out be inaccurate, incomplete, or fragmentary. Obviously, quality experimental directly depends on many factors: laboratory equipment, organization internal regulations, skills researchers, and so on. As a violations algorithms protocols initial streams processing – there are errors distortions data, that is why performing solid multistep data-curation process crucial for such procedures. Data curation procedure was performed approximately 60% rejected (due various errors, incomplete absent records physicochemical parameters conditions experiment), followed up by using zeta-potential value dataset 37 sizes nanoparticles 14 metal oxides calculation 1D SiRMS descriptors as well «liquid drop» model cross-descriptors. An efficient consensus built (R2 = 0.88, R2test 0.81). Predictive power 0.84) tested an external set 5 nano-oxides possibility satisfactory prediction shown. Prediction within domain applicability obtained QSPR confirmed Williams plot. The interpretation final carried it found contribution distributed between individual cross-descriptors 46% 54% respectively. 59%, second group 41% (liquid drop 29%, characterizing atom 12%). It most influential characteristics reflect nature oxides. electrostatic interactions have highest contribution.