Predicting chemical toxicity towards Raphidocelis subcapitata with quantum chemical descriptors

Abstract

The green alga Raphidocelis subcapitata is an important model organism in toxicity studies. This paper, for the first time, reported classification models for Raphidocelis subcapitata toxicities (pEC₅₀ and pEC₁₀), by applying 43 quantum chemical descriptors and the random forest algorithm. Both classification models for the toxicities pEC₅₀ and pEC₁₀ have prediction accuracy values of 100 % for the training set (251 organics) and above 90 % for the test set (83 organics). The quantum chemical descriptors selected include atomic charges, multipole moments, frontier orbital energies and kinetic energies, polarizability and thermochemistry properties. They are related to reaction sites, electrophilic and nucleophilic reactivity, chemical bond formation, hydrogen bonds and electrostatic force. The two classification models based on a sufficiently large sample set provide an important tool for assessing the toxicity categories of organics towards Raphidocelis subcapitata.