Polycyclic aromatic hydrocarbons (PAHs) represent a common group of environmental pollutants that endanger various aquatic organisms via various pathways. To better prioritize the ecotoxicological hazard of PAHs to aquatic environment, we used 2D descriptors-based quantitative structure-toxicity relationship (QSTR) to assess the toxicity of PAHs toward six aquatic model organisms spanning three trophic levels. According to strict OECD guideline, six easily interpretable, transferable and reproducible 2D-QSTR models were constructed with high robustness and reliability. A mechanistic interpretation unveiled the key structural factors primarily responsible for controlling the aquatic ecotoxicity of PAHs. Furthermore, quantitative read-across and different machine learning approaches were employed to validate and optimize the modelling approach. Importantly, the optimum QSTR models were further applied for predicting the ecotoxicity of hundreds of untested/unknown PAHs gathered from Pesticide Properties Database (PPDB). Especially, we provided a priority list in terms of the toxicity of unknown PAHs to six aquatic species, along with the corresponding mechanistic interpretation. In summary, the models can serve as valuable tools for aquatic risk assessment and prioritization of untested or completely new PAHs chemicals, providing essential guidance for formulating regulatory policies.

With increasing annual chemical development and production, safety testing demands and requirements have also increased. In addition to traditional animal testing, quantitative structure-activity relationship (QSAR) modelling can be used to predict the biological effect of a chemical structure, based on the analysis of quantitative characteristics of structure features. Whilst suitable for e.g., pharmaceuticals, other compounds can be more challenging to model. The naturally occurring heavy metal mercury speciates in the environment, with some toxic species accumulating in aquatic organisms. Although this is well known, only little data is available from (eco)toxicological studies, none of which account for this speciation behaviour. The present work highlights the current toxicity data for mercury in aquatic animals and gaps in our understanding and data for future QSAR modelling. All publicly available ecotoxicology data was obtained from databases and literature. Only few studies could be determined that assessed mercury toxicity in aquatic species. Of these, likely speciation products were determined using PHREEQc. This highlighted that the mercury exposure species was not always the predominant species in the medium. Finally, the descriptors for the modelled species were obtained from ChemDes, highlighting the limited availability of such details. Additional testing is required, accounting for speciation and biological interactions, to successfully determine the toxicity profile of different mercury species in aquatic environments. In the present work, insufficient mercury-species specific data was obtained, to conduct QSAR modelling successfully. This highlights a significant lack of data, for a heavy metal with potentially fatal repercussions.

Fused and non-fused polycyclic aromatic hydrocarbons (FNFPAHs) are a type of organic compounds widely occurring in the environment that pose a potential hazard to ecosystem and public health, and thus receive extensive attention from various regulatory agencies. Here, quantitative structure-activity relationship (QSAR) models were constructed to model the ecotoxicity of FNFPAHs against two aquatic species, Daphnia magna and Oncorhynchus mykiss. According to the stringent OECD guidelines, we used genetic algorithm (GA) plus multiple linear regression (MLR) approach to establish QSAR models of the two aquatic toxicity endpoints: D. magna (48 h LC50) and O. mykiss (96 h LC50). The models were established using simple 2D descriptors with explicit physicochemical significance and evaluated using various internal/external validation metrics. The results clearly show that both models are statistically robust (QLOO2 = 0.7834 for D. magna and QLOO2 = 0.8162 for O. mykiss), have good internal fitness (R2 = 0.8159 for D. magna and R2 = 0.8626 for O. mykiss and external predictive ability (D. magna: Rtest2 = 0.8259, QFn2 = 0.7640∼0.8140, CCCtest = 0.8972; O. mykiss:Rtest2 = 0.8077, QFn2 = 0.7615∼0.7722, CCCtest = 0.8910). To prove the predictive performance of the developed models, an additional comparison with the standard ECOSAR tool obviously shows that our models have lower RMSE values. Subsequently, we utilized the best models to predict the true external set compounds collected from the PPDB database to further fill the toxicity data gap. In addition, consensus models (CMs) that integrate all validated individual models (IMs) were more externally predictive than IMs, of which CM2 has the best prediction performance towards the two aquatic species. Overall, the models presented here could be used to evaluate unknown FNFPAHs inside the domain of applicability (AD), thus being very important for environmental risk assessment under current regulatory frameworks.

In this study, photodegradation experiments simulating the exposure conditions of sunlight on the commonly detected in surface and wastewater contaminants atorvastatin (ATV), bezafibrate (BEZ), oxybenzone (OXZ), and tris(2-butoxyethyl)phosphate (TBEP) were conducted as the fate of these compounds and their transformation products (TPs) was followed. Then a nontargeted analysis was carried out on an urban river to confirm the environmental occurrence of the TPs after which the ECOSAR software was used to generate predicted effect levels of toxicity of the detected TPs on aquatic organisms. Five TPs of ATV were tentatively identified including two stable ones at the end of the experiment: ATV_TP557a and ATV_TP575, that were the product of hydroxylation. Complete degradation of OXZ was observed in the experiment with no significant TP identified. BEZ remained stable and largely undegraded at the end of the exposure. Five TPs of TBEP were found including four that were stable at the end of the experiment: TBEP_TP413, TBEP_TP415, TBEP_TP429, and TBEP_TP343. In the nontargeted analysis, ATV_TP557b, a positional isomer of ATV_TP557a, ATV_TP575 and the 5 TPs of TBEP were tentatively identified. The predicted concentration for effect levels were lower for ATV_TP557b compared to ATV indicating the TP is potentially more toxic than the parent compound. All the TPs of TBEP showed lower predicted toxicity toward aquatic organisms than their parent compound. These results highlight the importance of conducting complete workflows from laboratory experiments, followed by nontargeted analysis to confirm environmental occurrence to end with predicted toxicity to better communicate concern of the newfound TPs to monitoring programs.

Accurate in silico predictions of chemical substance ecotoxicity has become an important issue in recent years. Most conventional methods, such as the Ecological Structure-Activity Relationship (ECOSAR) model, cluster chemical substances empirically based on structural information and then predict toxicity by employing a log P linear regression model. Due to empirical classification, the prediction accuracy does not improve even if new ecotoxicity test data are added. In addition, most of the conventional methods are not appropriate for predicting the ecotoxicity on inorganic and/or ionized compounds. Furthermore, a user faces difficulty in handling multiple Quantitative Structure-Activity Relationship (QSAR) formulas with one chemical substance. To overcome the flaws of the conventional methods, in this study a new method was developed that applied unsupervised machine learning and graph theory to predict acute ecotoxicity. The proposed machine learning technique is based on the large AIST-MeRAM ecotoxicity test dataset, a software program developed by the National Institute of Advanced Industry Science and Technology for Multi-purpose Ecological Risk Assessment and Management, and the Molecular ACCess System (MACCS) keys that vectorize a chemical structure to 166-bit binary information. The acute toxicity of fish, daphnids, and algae can be predicted with good accuracy, without requiring log P and linear regression models in existing methods. Results from the new method were cross-validated and compared with ECOSAR predictions and show that the new method provides better accuracy for a wider range of chemical substances, including inorganic and ionized compounds.