Using quasi-SMILES for the predictive modeling of the safety of 574 metal oxide nanoparticles measured in different experimental conditions

De novo Design of Biocompatible Nanomaterials Using Quasi-SMILES and Recurrent Neural Networks

Screening nanomaterials (NMs) with desired properties from the extensive chemical space presents significant challenges. The potential toxicity of NMs further limits their applications in biological systems. Traditional methods struggle with these complexities, but generative models offer a possible solution to producing new molecules without prior knowledge. However, converting complex 3D nanostructures into computer-readable formats remains a critical prerequisite. To overcome these challenges, we proposed an innovative deep-learning framework for the de novo design of biocompatible NMs. This framework comprises two predictive models and a generative model, utilizing a Quasi-SMILES representation to encode three-dimensional structural information on NMs. Our generative model successfully created 289 new NMs not previously seen in the training set. The predictive models identified a particularly promising NM characterized by high cellular uptake and low toxicity. This NM was successfully synthesized, and its predicted properties were experimentally validated. Our approach advances the application of artificial intelligence in NM design and provides a practical solution for balancing functionality and toxicity in NMs.

A review on the structural characterization of nanomaterials for nano-QSAR models

Quantitative structure-activity relationship (QSAR) models are routinely used to predict the properties and biological activity of chemicals to direct synthetic advances, perform massive screenings, and even to register new substances according to international regulations. Currently, nanoscale QSAR (nano-QSAR) models, adapting this methodology to predict the intrinsic features of nanomaterials (NMs) and quantitatively assess their risks, are blooming. One of the challenges is the characterization of the NMs. This cannot be done with a simple SMILES representation, as for organic molecules, because their chemical structure is complex, including several layers and many inorganic materials, and their size and geometry are key features. In this review, we survey the literature for existing predictive models for NMs and discuss the variety of calculated and experimental features used to define and describe NMs. In the light of this research, we propose a classification of the descriptors including those that directly describe a component of the nanoform (core, surface, or structure) and also experimental features (related to the nanomaterial's behavior, preparation, or test conditions) that indirectly reflect its structure.

Building species trait-specific nano-QSARs: Model stacking, navigating model uncertainties and limitations, and the effect of dataset size

A strong need exists for broadly applicable nano-QSARs, capable of predicting toxicological outcomes towards untested species and nanomaterials, under different environmental conditions. Existing nano-QSARs are generally limited to only a few species but the inclusion of species characteristics into models can aid in making them applicable to multiple species, even when toxicity data is not available for biological species. Species traits were used to create classification- and regression machine learning models to predict acute toxicity towards aquatic species for metallic nanomaterials. Afterwards, the individual classification- and regression models were stacked into a meta-model to improve performance. Additionally, the uncertainty and limitations of the models were assessed in detail (beyond the OECD principles) and it was investigated whether models would benefit from the addition of more data. Results showed a significant improvement in model performance following model stacking. Investigation of model uncertainties and limitations highlighted the discrepancy between the applicability domain and accuracy of predictions. Data points outside of the assessed chemical space did not have higher likelihoods of generating inadequate predictions or vice versa. It is therefore concluded that the applicability domain does not give complete insight into the uncertainty of predictions and instead the generation of prediction intervals can help in this regard. Furthermore, results indicated that an increase of the dataset size did not improve model performance. This implies that larger dataset sizes may not necessarily improve model performance while in turn also meaning that large datasets are not necessarily required for prediction of acute toxicity with nano-QSARs.

Nanoparticles and neurodegeneration: Insights on multiple pathways of programmed cell death regulated by nanoparticles

Currently, nanoparticles (NPs) are extensively applied in the diagnosis and treatment of neurodegenerative diseases (NDs). With the rapid development and increasing exposure to the public, the potential neurotoxicity associated with NDs caused by NPs has attracted the researchers' attentions but their biosafety assessments are still far behind relevant application studies. Based on recent research, this review aims to conduct a comprehensive and systematic analysis of neurotoxicity induced by NPs. The 191 studies selected according to inclusion and exclusion criteria were imported into the software, and the co-citations and keywords of the included literatures were analyzed to find the breakthrough point of previous studies. According to the available studies, the routes of NPs entering into the normal and injured brain were various, and then to be distributed and accumulated in living bodies. When analyzing the adverse effects induced by NPs, we focused on multiple programmed cell deaths (PCDs), especially ferroptosis triggered by NPs and their tight connection and crosstalk that have been found playing critical roles in the pathogenesis of NDs and their underlying toxic mechanisms. The activation of multiple PCD pathways by NPs provides a scientific basis for the occurrence and development of NDs. Furthermore, the adoption of new methodologies for evaluating the biosafety of NPs would benefit the next generation risk assessment (NGRA) of NPs and their toxic interventions. This would help ensure their safe application and sustainable development in the field of medical neurobiology.

Quasi-SMILES as a basis to build up models of endpoints for nanomaterials

Simplified molecular input-line entry system (SMILES) is a format for representing of the molecular structure. Quasi-SMILES is an extended format for representing molecular structure data and some eclectic data, which in principle could be applied to improve a model's predictive potential. Nano-quantitative structure-property relationships (nano-QSPRs) for energy gap (Eg, eV) of the metals oxide nanoparticles based on the quasi-SMILES give a predictive model for Eg, characterized by the following statistical quality for external validation set n = 22, R2 = 0.83, RMSE = 0.267.

Using Machine Learning to Predict Adverse Effects of Metallic Nanomaterials to Various Aquatic Organisms

The wide production and use of metallic nanomaterials (MNMs) leads to increased emissions into the aquatic environments and induces high potential risks. Experimentally evaluating the (eco)toxicity of MNMs is time-consuming and expensive due to the multiple environmental factors, the complexity of material properties, and the species diversity. Machine learning (ML) models provide an option to deal with heterogeneous data sets and complex relationships. The present study established an in silico model based on a machine learning properties-environmental conditions-multi species-toxicity prediction model (ML-PEMST) that can be applied to predict the toxicity of different MNMs toward multiple aquatic species. Feature importance and interaction analysis based on the random forest method indicated that exposure duration, illumination, primary size, and hydrodynamic diameter were the main factors affecting the ecotoxicity of MNMs to a variety of aquatic organisms. Illumination was demonstrated to have the most interaction with the other features. Moreover, incorporating additional detailed information on the ecological traits of the test species will allow us to further optimize and improve the predictive performance of the model. This study provides a new approach for ecotoxicity predictions for organisms in the aquatic environment and will help us to further explore exposure pathways and the risk assessment of MNMs.

Exploring the potential of in silico machine learning tools for the prediction of acute Daphnia magna nanotoxicity

Engineered nanomaterials (ENMs) are ubiquitous nowadays, finding their application in different fields of technology and various consumer products. Virtually any chemical can be manipulated at the nano-scale to display unique characteristics which makes them appealing over larger sized materials. As the production and development of ENMs have increased considerably over time, so too have concerns regarding their adverse effects and environmental impacts. It is unfeasible to assess the risks associated with every single ENM through in vivo or in vitro experiments. As an alternative, in silico methods can be employed to evaluate ENMs. To perform such an evaluation, we collected data from databases and literature to create classification models based on machine learning algorithms in accordance with the principles laid out by the OECD for the creation of QSARs. The aim was to investigate the performance of various machine learning algorithms towards predicting a well-defined in vivo toxicity endpoint (Daphnia magna immobilization) and also to identify which features are important drivers of D. magna in vivo nanotoxicity. Results indicated highly comparable model performance between all algorithms and predictive performance exceeding ∼0.7 for all evaluated metrics (e.g. accuracy, sensitivity, specificity, balanced accuracy, Matthews correlation coefficient, area under the receiver operator characteristic curve). The random forest, artificial neural network, and k-nearest neighbor models displayed the best performance but this was only marginally better compared to the other models. Furthermore, the variable importance analysis indicated that molecular descriptors and physicochemical properties were generally important within most models, while features related to the exposure conditions produced slightly conflicting results. Lastly, results also indicate that reliable and robust machine learning models can be generated for in vivo endpoints with smaller datasets.

Prediction models on biomass and yield of rice affected by metal (oxide) nanoparticles using nano-specific descriptors

The use of in silico tools to investigate the interactions between metal (oxide) nanoparticles (NPs) and plant biological responses is preferred because it allows us to understand molecular mechanisms and improve prediction efficiency by saving time, labor, and cost. In this study, four models (C5.0 decision tree, discriminant function analysis, random forest, and stepwise multiple linear regression analysis) were applied to predict the effect of NPs on rice biomass and yield. Nano-specific descriptors (size-dependent molecular descriptors and image-based descriptors) were introduced to estimate the behavior of NPs in plants to appropriately represent the wide space of NPs. The results showed that size-dependent molecular descriptors (e.g., E-state and connectivity indices) and image-based descriptors (e.g., extension, area, and minimum ferret diameter) were associated with the behavior of NPs in rice. The performance of the constructed models was within acceptable ranges (correlation coefficient ranged from 0.752 to 0.847 for biomass and from 0.803 to 0.905 for yield, while the accuracy ranged from 64% to 77% for biomass and 81% to 89% for yield). The developed model can be used to quickly and efficiently evaluate the impact of NPs under a wide range of experimental conditions and sufficient training data.

Nano-QSAR modeling for predicting the cytotoxicity of metallic and metal oxide nanoparticles: A review

Given the rapid development of nanotechnology, it is crucial to understand the effects of nanoparticles on living organisms. However, it is laborious to perform toxicological tests on a case-by-case basis. Quantitative structure-activity relationship (QSAR) is an effective computational technique because it saves time, costs, and animal sacrifice. Therefore, this review presents general procedures for the construction and application of nano-QSAR models of metal-based and metal-oxide nanoparticles (MBNPs and MONPs). We also provide an overview of available databases and common algorithms. The molecular descriptors and their roles in the toxicological interpretation of MBNPs and MONPs are systematically reviewed and the future of nano-QSAR is discussed. Finally, we address the growing demand for novel nano-specific descriptors, new computational strategies to address the data shortage, in situ data for regulatory concerns, a better understanding of the physicochemical properties of NPs with bioactivity, and, most importantly, the design of nano-QSAR for real-life environmental predictions rather than laboratory simulations.

CORAL: Development of a hybrid descriptor based QSTR model to predict the toxicity of dioxins and dioxin-like compounds with correlation intensity index and consensus modelling

In the present study, ninety-five halogenated dioxins and related chemicals (dibenzo-p-dioxins, dibenzofurans, biphenyls, and naphthalene) with endpoint pEC₅₀ were used to develop twelve quantitative structure toxicity relationship (QSTR) models using inbuilt Monte Carlo algorithm of CORAL software. The hybrid optimal descriptor of correlation weights (DCW) using a combination of SMILES and HSG (hydrogen suppressed graph) was employed to generate QSTR models. Three target functions i.e. TF₁ (W_IIC=W_CII=0), TF₂ (W_IIC= 0.3 & W_CII=0) and TF₃ (W_IIC= 0.0 &W_CII=0.3) were employed to develop robust QSTR models and the statistical outcomes of each target function were compared with each other. The correlation intensity index (CII) was found a reliable benchmark of the predictive potential for QSTR models. The numerical value of the determination coefficient of the validation set of split 1 computed by TF₃ was found highest (R_Valid²=0.8438). The fragments responsible for the toxicity of dioxins and related chemicals were also identified in terms of the promoter of increase/decrease for pEC₅₀. Three random splits (Split 1, Split 2 and Split 4) were selected for the extraction of the promoter of increase/decrease for pEC₅₀. In the last, consensus modelling was performed using the intelligent consensus tool of DTC lab (https://dtclab.webs.com/software-tools). The original consensus model, which was created by combining four distinct models employing the split 4 arrangement, was more predictive for the validation set and the numerical value of the determination coefficient of the test set (validation set) was increased from 0.8133 to 0.9725. For the validation set of split 4, the mean absolute error (MAE 100%) was also lowered from 0.513 to 0.2739.

Nanosafety: An Evolving Concept to Bring the Safest Possible Nanomaterials to Society and Environment

The use of nanomaterials has been increasing in recent times, and they are widely used in industries such as cosmetics, drugs, food, water treatment, and agriculture. The rapid development of new nanomaterials demands a set of approaches to evaluate the potential toxicity and risks related to them. In this regard, nanosafety has been using and adapting already existing methods (toxicological approach), but the unique characteristics of nanomaterials demand new approaches (nanotoxicology) to fully understand the potential toxicity, immunotoxicity, and (epi)genotoxicity. In addition, new technologies, such as organs-on-chips and sophisticated sensors, are under development and/or adaptation. All the information generated is used to develop new in silico approaches trying to predict the potential effects of newly developed materials. The overall evaluation of nanomaterials from their production to their final disposal chain is completed using the life cycle assessment (LCA), which is becoming an important element of nanosafety considering sustainability and environmental impact. In this review, we give an overview of all these elements of nanosafety.

Health and environmental safety of nanomaterials: O Data, Where Art Thou?

Nanotechnology keeps drawing attention due to the great tunable properties of nanomaterials in comparison to their bulk conventional materials. The growth of nanotechnology in combination with the digitization era has led to an increased need of safety related data. In addition to safety, new data-driven paradigms on safe and sustainable by design materials are stressing the necessity of data even more. Data is a fundamental asset to the scientific community in studying and analysing the entire life-cycle of nanomaterials. Unfortunately, data exist in a scattered fashion, in different sources and formats. To our knowledge, there is no study focusing on aspects of actual data-structure knowledge that exists in literature and databases. The purpose of this review research is to transparently and comprehensively, display to the nanoscience community the datasets readily available for machine learning purposes making it convenient and more efficient for the next users such as modellers or data curators to retrieve information. We systematically recorded the features and descriptors available in the datasets and provide synopsised information on their ranges, forms and metrics in the supplementary material.

Unswerving modeling of hepatotoxicity of cadmium containing quantum dots using amalgamation of quasiSMILES, index of ideality of correlation, and consensus modeling

Liver toxicity of quantum dots varies with size, concentration, and other structural as well as experimental parameters. For modeling hepatotoxicity, the eclectic data associated with cadmium containing quantum dots have been used in the creation of quasiSMILES for their representation. The core diameter is normalized for wider applicability and the index of the ideality of correlation is applied to construct the better quantitative features toxicity relationship models. Total eight splits are created and the best model is obtained through split 1 with better prediction criteria of validation set objects. The values of all statistical criteria used in the quality determination of a QSAR model are within the specified range for all the eight toxicity models developed here. Factors like TGA ligand and 0.6-0.7 nm diameter are favorable for liver toxicity while L-cysteine ligand and 0.5-0.6 nm core diameter are helpful in the reduction of toxicity. Further, the intelligent consensus modeling process forms a total of 40 individual and 20 consensus models and the best individual and consensus models are 'Good' in MAE-based criteria. The consensus modeling enhances the prediction ability as well as the accuracy of the developed models and increases the applicability space of the built models for hepatotoxicity prediction of quantum dots.

Prediction of power conversion efficiency of phenothiazine-based dye-sensitized solar cells using Monte Carlo method with index of ideality of correlation

Simplified molecular-input line-entry system (SMILES) notation and inbuilt Monte Carlo algorithm of CORAL software were employed to construct generative and prediction QSPR models for the analysis of the power conversion efficiency (PCE) of 215 phenothiazine derivatives. The dataset was divided into four splits and each split was further divided into four sets. A hybrid descriptor, a combination of SMILES and hydrogen suppressed graph (HSG), was employed to build reliable and robust QSPR models. The role of the index of ideality of correlation (IIC) was also studied in depth. We performed a comparative study to predict PCE using two target functions (TF₁ without IIC and TF₂ with IIC). Eight QSPR models were developed and the models developed with TF₂ was shown robust and reliable. The QSPR model generated from split 4 was considered a leading model. The different statistical benchmarks were computed for the lead model and these were rtraining set2=0.7784; rinvisible training set2=0.7955; rcalibration set2=0.7738; rvalidation set2=0.7506; Qtraining set2=0.7691; Qinvisible training set2=0.7850; Qcalibration set2=0.7501;  Qvalidation set2=0.7085; IIC_{training set} = 0.8590; IIC_{invisible training set} = 0.8297; IIC_{calibration set} = 0.8796; IIC_{validation set} = 0.8293, etc. The promoters of increase and decrease of endpoint PCE were also extracted.