Predictive models in ecotoxicology: Bridging the gap between scientific progress and regulatory applicability-Remarks and research needs

Explainable machine learning models for predicting the acute toxicity of pesticides to sheepshead minnow (Cyprinodon variegatus)

A quantitative structure-activity relationship (QSAR) study was conducted on 313 pesticides to predict their acute toxicity to Sheepshead minnow (Cyprinodon variegatus) by using DRAGON descriptors. Essentials accounting for a reliable model were all considered carefully, giving full consideration to the OECD (Organization for Economic Co-operation and Development) principles for QSAR acceptability in regulation during the model construction and assessment process. Nine variables were selected through the forward stepwise regression method and used as inputs to construct both linear and nonlinear models. The obtained models were validated internally and externally. Generally, machine learning-based methods, namely support vector machine (SVM), random forest (RF), and projection pursuit regression (PPR), perform better than the multiple linear regression (MLR) model. The statistical results (R2 = 0.682-0.933, Q2LOO = 0.604-0.659, Q2F1 = 0.740-0.796, CCC = 0.861-0.882) of the developed models show that they are robust, reliable, reproducible, accurate and predictive. Comparatively, the RF model performs best, giving predictive correlation coefficient Q2 of 0.814, root mean squared error (RMSE) of 0.658 and mean absolute error (MAE) of 0.534 for the test set, respectively. The RF model (as well as SVM and PPR models) was visualized and explained by using the SHapley Additive explanation (SHAP) analysis to enhance its transparency and credibility. In addition, the applicability domain (AD) range of the RF model was characterized by the Williams plot and the tree manifold approximation and projection (TMAP) technology was utilized to illustrate similarity and diversity of the entire data space, to assist in the analysis of the outliers. Activity cliff detection was investigated by using Arithmetic Residuals in K-groups Analysis (ARKA) descriptors. It was found that none of the pesticides was identified as an activity cliff in the training set or a potential prediction cliff in the test set. Therefore, the RF model fulfills each OECD principle in regulation for QSAR models. The research in this work will aid in the in silico QSAR prediction of the acute toxicity to Sheepshead minnow (Cyprinodon variegatus) for untested and new toxic pesticides and can also be extended to other studies.

Analysis of implicit and explicit uncertainties in QSAR prediction of chemical toxicity: A case study of neurotoxicity

Although uncertainties expressed in texts within QSAR studies can guide quantitative uncertainty estimations, they are often overlooked during uncertainty analysis. Using neurotoxicity as an example, this study developed a method to support analysis of implicitly and explicitly expressed uncertainties in QSAR modeling studies. Text content analysis was employed to identify implicit and explicit uncertainty indicators, whereafter uncertainties within the indicator-containing sentences were identified and systematically categorized according to 20 uncertainty sources. Our results show that implicit uncertainty was more frequent within most uncertainty sources (13/20), while explicit uncertainty was more frequent in only three sources, indicating that uncertainty is predominantly expressed implicitly in the field. The most highly cited sources included Mechanistic plausibility, Model relevance and Model performance, suggesting they constitute sources of most concern. The fact that other sources like Data balance were not mentioned, although it is recognized in the broader QSAR literature as an area of concern, demonstrates that the output from the type of analysis conducted here must be interpreted in the context of the broader QSAR literature before conclusions are drawn. Overall, the method established here can be applied in other QSAR modeling contexts and ultimately guide efforts targeted towards addressing the identified uncertainty sources.

Mortality Pattern of Poecilus cupreus Beetles after Repeated Topical Exposure to Insecticide─Stochastic Death or Individual Tolerance?

The mortality of organisms exposed to toxicants has been attributed to either stochastic processes or individual tolerance (IT), leading to the stochastic death (SD) and IT models. While the IT model follows the principles of natural selection, the relevance of the SD model has been debated. To clarify why the idea of stochastic mortality has found its way into ecotoxicology, we investigated the mortality of Poecilus cupreus (Linnaeus, 1758) beetles from pesticide-treated oilseed rape (OSR) fields and unsprayed meadows, subjected to repeated insecticide treatments. We analyzed the mortality with the Kaplan-Meier estimator and general unified threshold model for survival (GUTS), which integrates SD and IT assumptions. The beetles were exposed three times, ca. monthly, to the same dose of Proteus 110 OD insecticide containing thiacloprid and deltamethrin, commonly used in the OSR fields. Kaplan-Meier analysis showed that the mortality of beetles from meadows was much higher after the first treatment than after the next two, indicating the IT model. Beetles from the OSR displayed approximately constant mortality after the first and second treatments, consistent with the SD model. GUTS analysis did not conclusively identify the better model, with the IT being marginally better for beetles from meadows and the SD better for beetles from OSR fields.

Multi- and Trans-Generational Effects on Daphnia Magna of Chlorpyrifos Exposures

Chlorpyrifos, a broad-spectrum neurotoxic organophosphate insecticide, is subject to atmospheric and hydrolytic transport from application sites to aquatic ecosystems. Across the landscape, concentrations in surface water can vary spatially and temporally according to seasonal use practices. Standardized bioassays can provide a screening-level understanding of aquatic receptor acute and chronic toxicity. However, these bioassays do not address ecologically relevant exposure patterns that may impact fitness and survival within and across generations. The aim of the present study was to evaluate the utility of a second-tier, screening-level methodology employing Daphnia magna multi- and transgenerational bioassays spanning four generations to investigate the effect of variable chronic chlorpyrifos exposure. The multigenerational assay consisted of continuous chlorpyrifos exposure across four consecutive 21-day bioassays using progeny from the previous assay for each successive generation. In the transgenerational assay, only the parent (F0) generation was exposed. For both assays, survival and reproduction were assessed across treatments and generations. Results indicated that (1) following continuous chlorpyrifos exposure at ecologically relevant concentrations to four generations of D. magna, the highest treatment showed an apparent tolerance response for both survival and reproductive success in the F3 generation, and (2) chlorpyrifos exposure to the F0 generation did not result in treatment effects in the unexposed F1, F2, and F3 generations in the apical endpoints of survival and reproduction. Employing a suite of acute and chronic bioassays, including chronic exposures spanning multiple generations, allows for a more robust screening-level evaluation of the potential impact of chlorpyrifos on aquatic receptors for variable periods of exposure. Environ Toxicol Chem 2022;41:1054-1065. © 2021 SETAC.

Effect Modeling Quantifies the Difference Between the Toxicity of Average Pesticide Concentrations and Time-Variable Exposures from Water Quality Monitoring

Synthetic chemicals are frequently detected in water bodies, and their concentrations vary over time. Water monitoring programs typically employ either a sequence of grab samples or continuous sampling, followed by chemical analysis. Continuous time-proportional sampling yields the time-weighted average concentration, which is taken as proxy for the real, time-variable exposure. However, we do not know how much the toxicity of the average concentration differs from the toxicity of the corresponding fluctuating exposure profile. We used toxicokinetic-toxicodynamic models (invertebrates, fish) and population growth models (algae, duckweed) to calculate the margin of safety in moving time windows across measured aquatic concentration time series (7 pesticides) in 5 streams. A longer sampling period (14 d) for time-proportional sampling leads to more deviations from the real chemical stress than shorter sampling durations (3 d). The associated error is a factor of 4 or less in the margin of safety value toward underestimating and an error of factor 9 toward overestimating chemical stress in the most toxic time windows. Under- and overestimations occur with approximate equal frequency and are very small compared with the overall variation, which ranged from 0.027 to 2.4 × 10¹⁰ (margin of safety values). We conclude that continuous, time-proportional sampling for a period of 3 and 14 d for acute and chronic assessment, respectively, yields sufficiently accurate average concentrations to assess ecotoxicological effects. Environ Toxicol Chem 2020;39:2158-2168. © 2020 The Authors. Environmental Toxicology and Chemistry published by Wiley Periodicals LLC on behalf of SETAC.

Consensus versus Individual QSARs in Classification: Comparison on a Large-Scale Case Study

Consensus strategies have been widely applied in many different scientific fields, based on the assumption that the fusion of several sources of information increases the outcome reliability. Despite the widespread application of consensus approaches, their advantages in quantitative structure–activity relationship (QSAR) modeling have not been thoroughly evaluated, mainly due to the lack of appropriate large-scale data sets. In this study, we evaluated the advantages and drawbacks of consensus approaches compared to single classification QSAR models. To this end, we used a data set of three properties (androgen receptor binding, agonism, and antagonism) for approximately 4000 molecules with predictions performed by more than 20 QSAR models, made available in a large-scale collaborative project. The individual QSAR models were compared with two consensus approaches, majority voting and the Bayes consensus with discrete probability distributions, in both protective and nonprotective forms. Consensus strategies proved to be more accurate and to better cover the analyzed chemical space than individual QSARs on average, thus motivating their widespread application for property prediction. Scripts and data to reproduce the results of this study are available for download.