Developing predictive models for assessing LC50 of organic contaminants in Gammarus species using interpretable structural parameters

Gammarus species are crucial indicators of environmental health, making them important for ecological studies and water quality assessments. They offer a wide range of specific responses regarding the median lethal concentration (LC₅₀) of organic compounds. This research presents four predictive models to determine the LC₅₀ of chemicals impacting selected gammarid amphipods: Gammarus lacustris, Gammarus fasciatus, Gammarus pulex, and Gammarus pseudolimnaeus. These species are recognized for their sensitivity to various pollutants and are among the most sensitive aquatic invertebrates. The new models provide straightforward methods for estimating the pLC₅₀ (−log LC₅₀/molecular weight) of various organic compounds based on interpretable structural parameters including the number of effective functional groups, the types of atoms present, and various structural characteristics of organic molecules. This study aims to leverage the largest available experimental dataset compared to prior quantitative structure–activity relationship (QSAR) models for these gammarid amphipods. The dataset contained toxicity data for 91 compounds affecting Gammarus fasciatus, 50 for Gammarus lacustris, and 48 each for Gammarus pseudolimnaeus and Gammarus pulex, aligning with comparative QSAR models. External datasets included 13 compounds for Gammarus fasciatus, 2 for Gammarus lacustris, and 6 for Gammarus pseudolimnaeus. Efforts focus on using interpretable structural parameters of organic compounds rather than computer-based descriptors, as outlined in the existing QSAR models. For the species G. fasciatus, G. lacustris, G. pseudolimnaeus, and G. pulex, the R² ratios for the new models versus the best QSAR models are 0.915/0.728, 0.955/0.747, 0.976/0.769, and 0.970/0.768, respectively. The higher R² values in the new models demonstrate greater reliability and robustness in capturing the data's underlying relationships.