Predicting joint toxicity of chemicals by incorporating a weighted descriptor into a mixture model: Cases for binary antibiotics and binary nanoparticles

A prediction method that integrated a mixture descriptor with an established mixture toxicology method was proposed for the joint toxicity of chemical pollutants. A weighted descriptor derived from the single descriptor of each component was employed to calculate a mixture descriptor, which was successfully embedded into the generalized concentration addition (GCA) model named the extended GCA (XGCA) model. To develop and validate the proposed approach, binary antibiotic mixtures (ciprofloxacin and oxytetracycline) and metal-oxide (copper oxide and zinc oxide) nanoparticle mixtures were selected to study their toxicity to freshwater green algae. The results showed that concentration-response curve (CRC) derived from the XGCA model was closer to the observed CRC than those from the GCA, Concentration Addition (CA), and Independent Action (IA) models. The difference between effect concentrations predicted by the XGCA model and observed did not exceed a factor of 1.6. The XGCA model was relatively more accurate at predicting joint toxicity (in terms of effect concentra-tions and effect errors) than the reference models, independent of component types and mixture ratios. The XGCA model predicts the joint toxicity through molecular structural or nanostructural characters, thus modes of toxic action are not preconditions for predicting the toxicity of the mixtures. This result demonstrates the practicability of using the XGCA method in toxicity assessments of mixture pollutants with unknown modes of action.

Automated summary

A new method for predicting the joint toxicity of chemical pollutants was proposed, which integrated a mixture descriptor with an established mixture toxicology method. The results showed that this method provided more accurate predictions compared to traditional models.