BNNmix: A new approach for predicting the mixture toxicity of multiple components based on the back-propagation neural network

Prediction of toxicity and identification of key components for complex mixtures containing hormetic components

Mixtures containing hormetic components are likely to induce hormesis. However, due to the presence of stimulatory effects, predicting the toxicity of such mixtures and identifying their key components face challenges. This study investigated the complex relationship between the stimulatory effects of individual components and their mixtures, focusing on predicting mixture toxicity and identifying key components influencing this toxicity. Sixteen chemicals, commonly found in disinfectants and hand sanitizers, were selected to construct a complex mixture system containing hormetic components. Using Vibrio qinghaiensis sp.-Q67 as an indicator organism, the study employed microplate toxicity tests to collect toxicity data for individual chemicals and their mixtures. The independent action (IA) and back-propagation neural network (BPNN) methods were utilized to predict mixture toxicity, while global sensitivity analysis (GSA) identified key components affecting toxicity. Results revealed that six of the sixteen chemicals exhibited time-dependent hormesis. However, when combined into mixtures, the stimulatory effects observed in individual components tended to diminish or disappear, leading to higher overall toxicity, likely due to synergism. Traditional models like the IA significantly underestimated mixture toxicity, whereas the BPNN model demonstrated superior predictive performance. GSA identified five key components, and changes in the levels of some non-toxic components significantly altered the toxicity of the mixtures. Moreover, increasing the levels of certain key components could either increase or decrease the mixture's toxicity, making the strategy of reducing their concentration to control mixture toxicity ineffective. This study revealed the potential of neural networks in predicting the toxicity of mixtures containing hormetic components and the possible characteristics of the effects of key components on mixture toxicity.

Derivation of marine water quality criteria for copper based on artificial neural network model

The water chemical effects of copper have been a focus in the study of water quality criteria (WQC). Currently, multiple regression models are commonly used to quantitatively describe the impact of environmental factors on Cu toxicity in WQC studies. However, the influence of species-specific effects may consequently lead to poor prediction results of the regression models in practical application. For this issue, a backpropagation neural network (BPNN) model optimized using a genetic algorithm was developed in this study. The results showed when pooled data of given taxonomic groups were used, the BPNN mixed models had higher Adj.R2 for five out of seven groups in the predicted toxicity values compared to the MNLR mixed models. When using species-specific models, the BPNN model still showed higher predictive performance. Further comparison of the two models for the species M. galloprovincialis revealed that, in addition to the good predictive performance of the BPNN models, the pre-set species codes of different species in the taxonomic group for the BPNN mixed model also reduced the impact of species-specific effects among species. Finally, the WQCs under different water quality parameter ranges were obtained using predicted toxicity values from mixed BPNN and MNLR models. The short-term WQC range for common water quality parameters (salinity: 25-30 ppt, DOC: 0.5-2.5 mg/L) obtained from the BPNN mixed model in natural marine environments was 1.6-4.41 μg/L, which aligns with guidance values provided by major global institutions, demonstrating the feasibility of applying the BPNN mixed model to WQC derivation. This study aims to provide valuable references for future research on WQC.

Benzo(a)anthracene Targeting SLC1A5 to Synergistically Enhance PAH Mixture Toxicity

Human exposure to polycyclic aromatic hydrocarbons (PAHs) as mutagenic and carcinogenic pollutants in the environment often occurs in the form of mixtures. Although the mixture effects of PAHs have been previously recognized, the toxicological mechanisms to explain them still remain quite unclear. This study combined metabolomics and chemical proteomics methods to comprehensively understand the mixture effects of a PAH mixture including benzo(a)anthracene (BaA), benzo(b)fluoranthene (BbF), benzo(a)pyrene (BaP), and chrysene (CHR). Among them, BaA has shown a strong synergistic effect with other PAHs. Interestingly, BaA alone is not a potent oxidative stress inducer in liver cells but dose-dependently amplifies oxidative damage caused by the PAH mixture. Global metabolomics analysis results revealed damage to the antioxidant glutathione synthesis, which was caused by the glutamine depletion caused by BaA in the mixture. Subsequently, the label-free chemical proteomics and cellular thermal shift analysis (CETSA) demonstrated that the PAH mixture altered the thermal shift of glutamine transporter SLC1A5. Furthermore, Western blotting and the isothermal titration calorimetry (ITC) interaction measurements showed nanomolar KD values between BaA and SLC1A5. Overall, this study showed that BaA synergistically contributed to PAH mixture induced oxidative damage by targeting SLC1A5 to inhibit glutamate transport into cells, resulting in the inhibition of glutathione synthesis.

Innovative graph neural network approach for predicting soil heavy metal pollution in the Pearl River Basin, China

Predicting soil heavy metal (HM) content is crucial for monitoring soil quality and ensuring ecological health. However, existing methods often neglect the spatial dependency of data. To address this gap, our study introduces a novel graph neural network (GNN) model, Multi-Scale Attention-based Graph Neural Network for Heavy Metal Prediction (MSA-GNN-HMP). The model integrates multi-scale graph convolutional network (MS-GCN) and attention-based GNN (AGNN) to capture spatial relationships. Using surface soil samples from the Pearl River Basin, we evaluate the MSA-GNN-HMP model against four other models. The experimental results show that the MSA-GNN-HMP model has the best predictive performance for Cd and Pb, with a coefficient of determination (R2) of 0.841 for Cd and 0.886 for Pb, and the lowest mean absolute error (MAE) of 0.403 mg kg-1 for Cd and 0.670 mg kg-1 for Pb, as well as the lowest root mean square error (RMSE) of 0.563 mg kg-1for Cd and 0.898 mg kg-1 for Pb. In feature importance analysis, latitude and longitude emerged as key factors influencing the heavy metal content. The spatial distribution prediction trend of heavy metal elements by different prediction methods is basically consistent, with the high-value areas of Cd and Pb respectively distributed in the northwest and northeast of the basin center. However, the MSA-GNN-HMP model demonstrates superior detail representation in spatial prediction. MSA-GNN-HMP model has excellent spatial information representation capabilities and can more accurately predict heavy metal content and spatial distribution, providing a new theoretical basis for monitoring, assessing, and managing soil pollution.

An artificial intelligence-based model for predicting reproductive toxicity of bisphenol analogues mixtures to the rotifer Brachionus calyciflorus

The joint toxicity effects of mixtures, particularly reproductive toxicity, one of the main causes of aquatic ecosystem degradation, are often overlooked as it is impractical to test all mixtures. This study developed and evaluated the following models to predict the concentration response curve concerning the joint reproductive toxicity of mixtures of three bisphenol analogues (BPA, BPF, BPAF) on the rotifer Brachionus calyciflorus: concentration addition (CA), independent action (IA), and two deep neural network (DNN) models. One applied mixture molecular descriptors as input variables (DNN-QSAR), while the other applied the ratios of chemicals in the mixtures (DNN-Ratio). Descriptors related to molecular mass were found to be of greater importance and exhibited a proportional relationship with toxic effects. The results indicate that the range of correlation coefficients (R2) between predicted and measured values for various mixture rays by CA and IA models is 0.372 to 0.974 and - 0.970 to 0.586, respectively. The R2 values for DNN-Ratio and DNN-QSAR were 0.841 to 0.984 and 0.834 to 0.991, respectively, demonstrating that models developed by DNN significantly outperform traditional models in predicting the joint toxicity of mixtures. Furthermore, DNN-QSAR not only predicts mixture toxicity but also provides accurate toxicity predictions for BPA, BPF, and BPAF, with R2 values of 0.990, 0.616, and 0.887, respectively, while DNN-Ratio yields values of 0.920, 0.355, and - 0.495. The study also found that the joint effects of mixtures are primarily influenced by the total concentration of the mixtures, and an increase in total concentration shifts the joint effects towards addition. This study introduces a novel approach to predict joint toxicity and analyze the influencing factors of joint effects, providing a more comprehensive assessment of the ecological risk posed by mixtures.

Combined interaction of fungicides binary mixtures: experimental study and machine learning-driven QSAR modeling

Fungicide mixtures are an effective strategy in delaying the development of fungicide resistance. In this research, a fixed ratio ray design method was used to generate fifty binary mixtures of five fungicides with diverse modes of action. The interaction of these mixtures was then analyzed using CA and IA models. QSAR modeling was conducted to assess their fungicidal activity through multiple linear regression (MLR), support vector machine (SVM), and artificial neural network (ANN). Most mixtures exhibited additive interaction, with the CA model proving more accurate than the IA model in predicting fungicidal activity. The MLR model showed a good linear correlation between selected theoretical descriptors by the genetic algorithm and fungicidal activity. However, both ML-based models demonstrated better predictive performance than the MLR model. The ANN model showed slightly better predictability than the SVM model, with R2 and R2cv at 0.91 and 0.81, respectively. For external validation, the R2test value was 0.845. In contrast, the SVM model had values of 0.91, 0.78, and 0.77 for the same metrics. In conclusion, the proposed ML-based model can be a valuable tool for developing potent fungicidal mixtures to delay fungicidal resistance emergence.

Mixtures Recomposition by Neural Nets: A Multidisciplinary Overview

Artificial Neural Networks (ANNs) are transforming how we understand chemical mixtures, providing an expressive view of the chemical space and multiscale processes. Their hybridization with physical knowledge can bridge the gap between predictivity and understanding of the underlying processes. This overview explores recent progress in ANNs, particularly their potential in the 'recomposition' of chemical mixtures. Graph-based representations reveal patterns among mixture components, and deep learning models excel in capturing complexity and symmetries when compared to traditional Quantitative Structure-Property Relationship models. Key components, such as Hamiltonian networks and convolution operations, play a central role in representing multiscale mixtures. The integration of ANNs with Chemical Reaction Networks and Physics-Informed Neural Networks for inverse chemical kinetic problems is also examined. The combination of sensors with ANNs shows promise in optical and biomimetic applications. A common ground is identified in the context of statistical physics, where ANN-based methods iteratively adapt their models by blending their initial states with training data. The concept of mixture recomposition unveils a reciprocal inspiration between ANNs and reactive mixtures, highlighting learning behaviors influenced by the training environment.

How does ultrasound-assisted ionic liquid treatment affect protein? A comprehensive review of their potential mechanisms, safety evaluation, and physicochemical and functional properties

Proteins are essential to human health with enormous food applications. Despite their advantages, plant and animal proteins often exhibit limited molecular flexibility and poor solubility due to hydrogen bonds, hydrophobic interactions, and ionic interactions within their molecular structures. Thus, there is an urgent need to modify the rigid structure of proteins to enhance their stability and functional properties. Ultrasound-assisted ionic liquid (UA-IL) treatment for developing compound modification and producing proteins with excellent functional properties has received interest. However, no review specifically addresses the interactions between UA-ILs and proteins. Hence, this review focused on recent research advancements concerning the effects and potential reaction mechanisms of UA-ILs on the physicochemical properties (including particle size; primary, secondary, and tertiary structure; and surface morphology) as well as the functionality (such as solubility, emulsifying properties, and foaming ability) of proteins. Moreover, the safety evaluation of modified proteins was also discussed from various perspectives, such as acute and chronic toxicity, genotoxicity, cytotoxicity, and environmental and microbial toxicity. This review demonstrated that UA-IL treatment-induced protein structural changes significantly impact the functional characteristics of proteins. This treatment approach efficiently promotes protein structure stretching and spatial rearrangement through cavitation, thermal effects, and ionic interactions. As a result, the functional properties of modified proteins exhibited an obvious enhancement, thereby bringing more opportunities to utilize modified protein products in the food industry. Potential future directions for protein modification using UA-ILs were also proposed.

Combined effects of fluoroquinolone antibiotics and organophosphate flame retardants on Microcystis aeruginosa

As freshwater harmful algal blooms continue to rise in frequency and severity, increasing focus is made on the effects of mixed pollutants and the dominant cyanobacterial species Microcystis aeruginosa (M. aeruginosa). However, few studies have investigated whether M. aeruginosa has a synergistic relationship with two common pollutants, namely, organophosphate flame retardants (OPFRs) and fluoroquinolone antibiotics (FQs). In this paper, three FQs and three OPFRs commonly detected in freshwaters were selected to construct a ternary mixture of FQs, a ternary mixture of OPFRs, and a six-component mixture of OPFRs and FQs. The effects of single substance and mixture on the growth of M. aeruginosa were determined at 24, 48, 72, and 96 h, and the toxicities of the mixture were evaluated by concentration addition model and independent action model. The results showed that the mixture of FQs and the mixture of OPFRs do not show toxicological interaction. However, partial mixtures of OPFRs and FQs showed antagonism or synergism at different concentrations and times. This indicated that combined toxicities of OPFRs and FQs on M. aeruginosa were mixture ratio dependent, concentration dependent and time dependent. This study improves our understanding of the role of OPFRs and FQs in cyanobacterial outbreaks of Microcystis.

Environmental pollution impacts: Are p values over-valued?

An examination revealed the dominance of the published literature of environmental science by p values. Meanwhile, the use of effect size has been neglected in publications reporting primary data, yet the size of effect is often more informative than p values inference in assessing the effects of pollution on living organisms, comparing susceptibility/resistance among organisms, and ranking pollutants according to their potency, among others. Statistical significance does not necessarily mean biological, practical, or scientific significance, and its use based on (often misinterpreted) p values reflects the average response or effect at average conditions based on an assumed linear model fit to the entire sample. However, pollution impacts and organismal responses are rarely characterized by linear and symmetric features, and dichotomous 'statistical significance' based on p values is inadequate to fully describe data and findings. Considering 'the fallacy of the average', variance, and differential response of different population percentiles in new studies would provide otherwise wasted biologically, practically, or scientifically significant information. Since p values often inform as to whether some findings warrant further examination, journals should consider mandating the reporting of effect sizes and confidence intervals, together with p values (should they be used), to provide more integrated information regarding pollution impacts. Moreover, replacing 'statistical significance' with language of evidence, especially in key components of publications, such as abstracts and conclusions, could help preventing potential misleading of the public and decision and policy makers.

Zero-Energy Buildings and Energy Efficiency towards Sustainability: A Bibliometric Review and a Case Study

The rapid growth of sustainability has created a plethora of options for expanding zero-energy buildings (ZEBs) and energy efficiency in all aspects of life. In recent years, there has been a rise in interest in ZEBs, and many countries have adopted ZEBs as future energy targets to promote the sustainable development paradigm. The primary goal of this paper was to conduct a bibliometric review of current research on ZEBs and energy efficiency. The first part of this paper identifies new knowledge gaps as well as practical demands in the field of sustainable development. Furthermore, bibliometric analysis was performed using the Scopus database (i.e., 2592 articles) and a screening process was undertaken, with the result being 252 papers. This study draws attention to a body of knowledge by reviewing trends and patterns, major research topics, journals, countries, new approaches, emerging trends, and future directions for sustainable development. This study is unique in that it provides a comprehensive, updated review of ZEBs and energy efficiency trends. Moreover, this study could help identify limitations for future policymakers, practitioners, and academics. The empirical section of this paper, through a case study, presents an example of a low-energy single-family building located in Poland.

New methods of top-to-down mixture toxicity prediction: A case study of eliminating of the effects of cosolvent from binary mixtures

At present, the toxicity prediction of mixtures mainly focuses on the concentration addition (CA) and independent action (IA) based on individual toxicants to predict the toxicity of multicomponent mixtures. This process of predicting the toxicity of multicomponent mixtures based on single substances or low component mixtures is called down-to-top method in this study. However, due to the particularity of some toxicants, we have to use the top-to-down idea to obtain or eliminate the toxicity of some components from mixtures. For example, the toxicity of toxicants is obtained from the toxicity of a mixture with, especially toxic, cosolvent added. In the study, two top-to-down methods, the inverse CA (ICA) and inverse IA (IIA) models, were proposed to eliminate the effects of a certain component from multicomponent mixtures. Furthermore, taking the eight binary mixtures consisting of different shapes of cosolvents (isopropyl alcohol (IPA) having hormesis and dimethyl sulfoxide (DMSO)) and toxicants (two ionic liquids and two pesticides) as an example, combined with the interaction evaluated by CA and IA model, the influence of different shapes of components on top-to-down toxicity prediction was explored. The results showed that cosolvent IPA having hormesis may cause unpredictable effects, even at low concentrations, and should be used with caution. For DMSO, most of the toxicant's toxicity obtained by ICA and IIA models were almost in accordance with those observed experimentally, which showed that ICA and IIA could effectively eliminate the effects of cosolvent, even if toxic cosolvent, from the mixture. Ultimately, a frame of cosolvent use and toxicity correction for the hydrophobic toxicant were suggested based on the top-to-down toxicity prediction method. The proposed methods improve the existing framework of mixture toxicity prediction and provide a new idea for mixture toxicity evaluation and risk assessment.

Study of the joint action of multi-component mixtures based on parameter σ2(k∙ECx) characterizing the shape difference of concentration-response curves

A previous study has revealed that the parameter k∙EC_x, characterizing the shape of concentration-response curves (CRCs), could predict the combined toxicity of binary mixtures. This study further explored the predictability of multi-component mixtures. Eleven component mixtures were designed using the uniform design ray, and the acute toxicity of the eleven environmental pollutants and their mixtures to Vibrio fischeri was determined using microplate toxicity analysis. We used independent action (IA) and the effect residual ratio (ERR_x) models to evaluate the combined toxicity of multi-component mixtures and ascertain the functional relationship between σ²_(k∙ECx), a parameter characterizing the CRC morphological difference of multi-component mixtures, and combined toxicity. The variance σ²_(k∙ECx) of each component characteristic parameter of multi-component mixtures gradually increased in the concentration range, and the relationship between σ²_(k∙ECx) and ERR_x was consistent with the exponential function. The literature verification showed that this rule is generally applicable to the acute toxicity of multi-component mixtures to luminescent bacteria. The exponential function showed the variation rule of the joint action of multi-component mixtures. In the present study, the joint toxicity of multi-component mixtures can be predicted from single toxicity and small amount of multiple toxicity, circumventing complex multi-component toxicity experiments.

Study on the Combined Toxicities and Quantitative Characterization of Toxicity Sensitivities of Three Flavor Chemicals and Their Mixtures to Caenorhabditis elegans

It was shown that flavor chemicals with high toxicity sensitivities mean that small changes in their effective concentrations can lead to significant changes in toxicity. Flavors are widely used in personal care products. However, our study demonstrated that some flavor chemicals and their mixture rays have high toxicity sensitivities to Caenorhabditis elegans (C. elegans), which may have an impact on human health. In this paper, three flavor chemicals (benzyl alcohol, phenethyl alcohol, and cinnamaldehyde) were used as components of the mixture, and three binary mixture systems were constructed, respectively. Five mixture rays were designed for each mixture system by a direct equipartition ray design method. The lethal toxicities of the three flavor chemicals and mixture rays to C. elegans at three exposure volumes were determined. A new concept (inverse of the negative logarithmic concentration span (iSPAN)) was introduced to quantitatively evaluate the toxicity sensitivity of chemicals or mixture rays, and the combination index (CI) was employed to identify the toxicological interactions in the mixtures. It was shown that the three flavor chemicals as well as the binary mixture rays have a significant concentration-response relationship on the lethality of C. elegans. The iSPAN values of the three flavor chemicals and their mixture rays were larger than 3.000, showing very strong toxicity sensitivity to C. elegans. In mixture systems, the toxicity sensitivities of mixture rays with different mixture ratios were also different at different exposure volumes. In addition, it can be seen from the CI heat map that the toxicological interaction not only shows the mixture ratio dependence but also changes with the different exposure volumes, which implies that the mixtures consisting of flavor chemicals with high toxicity sensitivity have complex toxicological interactions. Therefore, in environmental risk assessment, special attention should be paid to chemicals with high toxicity sensitivities.

Mixture predicted no-effect concentrations derived by independent action model vs concentration addition model based on different species sensitivity distribution models

In the hazard assessment of mixtures, the mixture predicted no-effect concentration (mPNEC) is always derived by the concentration addition (CA) model (mPNEC_CA) to assess the risk of mixtures combined with exposure assessment. However, the independent action (IA) model, which is also widely used as the CA model in the prediction and evaluation of mixture toxicity, is always used to calculate the population fraction showing a predefined effect, not mPNEC, and this limits the application of IA model in the mixture risk assessment. In this study, we explored the process of mPNEC derived by the IA method (mPNEC_IA) based on the species sensitivity distribution (SSD) and compared mPNEC_IA with mPNEC_CA. Taking two common pesticides, dimethoate (DIM) and dichlorvos (DIC), exposed in the actual water environment as an example, their SSD models were constructed separately using nine distribution functions after toxicity data screening and quality testing. For both DIC and DIM, all different nine models had passed the Kolmogorov-Smirnov test. Then, the PNECs of two pesticides were derived based on SSD models. Finally, mPNEC_IA with different concentration ratios was derived and compared to mPNEC_CA based on 81 combinations of nine SSD models. Most mPNEC values derived by IA model were more conservative than those by CA. It is worth noting that the mPNEC_IA is more conservative than mPNEC_CA for the commonly used log-logit distribution (function 7), log-normal distribution (8), and log-Weibull distribution (9). This study provides a new direction for the application of IA in the risk assessment and enriches the framework of mixture risk assessment.

Combined lethal toxicities of pesticides with similar structures to Caenorhabditis elegans are not necessarily concentration additives

Studies have shown that the mixture toxicity of compounds with similar modes of action (MOAs) is usually predicted by the concentration addition (CA) model. However, due to the lack of toxicological information on compounds, more evidence is needed to determine whether the above conclusion is generally applicable. In general, the same type of compounds with similar chemical structures have similar MOAs, so whether the toxicities of the mixture of these compounds are additive needs to be further studied. In this paper, three types of pesticides with similar chemical structures (three organophosphoruses, two carbamates and two neonicotinoids) that may have similar MOAs were selected and five binary mixture systems were constructed. For each system, five mixture rays with different concentration ratios were designed by the direct equipartition ray design (EquRay) method. The mortality of Caenorhabditis elegans was regarded as the endpoint for the toxicity exposure to single pesticides and binary mixtures. The combined toxicities were evaluated simultaneously using the CA model, isobologram and combination index. The structural similarity of the same type of pesticides was quantitatively analyzed according to the MACCS molecular fingerprint and the slope of dose-response curve at pEC₅₀. The results show that the toxicities of neonicotinoid mixtures and carbamate mixtures are almost antagonistic. The entire mixture system of dichlorvos and dimethoate produced synergism, and four of the five mixture rays of dimethoate and methamidophos induced antagonism, while among the mixture rays of dichlorvos and methamidophos, different concentrations showed different interaction types. The results of structural similarity analysis show that the size of structural similarity showed a certain quantitative relationship with the toxicity interaction of mixtures, that is, the structural similarity of the same type of pesticides may show an additive action in a certain range.

US EPA: Is there room to open a new window for evaluating potential sub-threshold effects and ecological risks?

With a rule published on 6 January 2021, the US Environmental Protection Agency (EPA) considers for the first time sub-threshold responses, abandoning the use of default dose-response models. This may affect worldwide scientific research, in terms of research design and methodology, and regulatory actions in China and other countries.

Protein Model and Function Analysis in Quorum-Sensing Pathway of Vibrio qinghaiensis sp.-Q67

Bioluminescent bacteria are mainly found in marine habitats. Vibrio qinghaiensis sp.-Q67 (Q67), a nonpathogenic freshwater bacterium, has been a focus due to its wide use in the monitoring of environmental pollution and the assessment of toxicity. However, the lack of available crystal structures limits the elucidation of the structures of the functional proteins of the quorum-sensing (QS) system that regulates bacterial luminescence in Q67. In this study, 19 functional proteins were built through monomer and oligomer modeling based on their coding proteins in the QS system of Q67 using MODELLER. Except for the failure to construct LuxM due to the lack of a suitable template, 18 functional proteins were successfully constructed. Furthermore, the relationships between the function and predicted structures of 19 functional proteins were explored one by one according to the three functional classifications: autoinducer synthases and receptors, signal transmission proteins (phosphotransferases, an RNA chaperone, and a transcriptional regulator), and enzymes involved in bacterial bioluminescence reactions. This is the first analysis of the whole process of bioluminescence regulation from the perspective of nonpathogenic freshwater bacteria at the molecular level. It provides a theoretical basis for the explanation of applications of Q67 in which luminescent inhibition is used as the endpoint.