Computational simulation of bioaccumulation and trophic transfer of antibiotics mechanisms in aquatic food chain

Abstract

Numerous antibiotics have been detected in aquatic ecosystems and induced severe toxic effects on aquatic organisms. However, mechanisms of bioaccumulation and trophic transfer of antibiotics are not adequately discussed, to the best of our knowledge. In this context, the bidirectional selective effect values (BSEV) and trophic transfer efficiency ratio (TTER) of 24 antibiotics in a simulated food chain (Chlorella sorokiniana-Daphnia magna-Danio rerio) were first calculated to mirror the bioaccumulation and biomagnification. Based on estimates above, the multi-output machine learning (ML) models, including K nearest neighbor (KNN), Support vector machine (SVM), Extremely randomized trees (ERT) and Extreme gradient boosting (XGBoost), were constructed, followed by molecular dynamics (MD) simulation and density functional theory (DFT) calculation to explore the bioaccumulation and biomagnification mechanism. According to our results, sulfonamide antibiotics had greater capacity biomagnification, while β-lactam and tetracycline antibiotics showed opposite results. Meanwhile Cytochromes P450 (CYP450) in Danio rerio played a key role in the food chain. The ERT model exhibited reliable prediction with indicators of R² = 0.816, MAE = 0.039, MSE = 0.003, RMSE = 0.053 and MAPE = 8.923. The AATS5s was identified as the most contributing descriptor. The differences in the atomic composition, structure and binding ability to enzymes of antibiotics lead to the differences in their bioaccumulation. Van der Waals interactions (ΔE_vdw) and non-polar interactions (ΔG_nonpolar) were the main driving energy for the biometabolism capability of antibiotics. Tetracyclines are the most readily biometabolized, whereas sulfonamides are more difficult to biometabolize due to their low binding capacity and low reactivity.

Keywords: Antibiotic; Food chain; Machine learning model; Trophic transfer.