Simplified Molecular Input Line Entry Research Articles

ChemGenerator: a web server for generating potential ligands for specific targets.

In drug discovery, one of the most important tasks is to find novel and biologically active molecules. Given that only a tip of iceberg of drugs was founded in nearly one-century's experimental exploration, it shows great significance to use in silico methods to expand chemical database and profile drug-target linkages. In this study, a web server named ChemGenerator was proposed to generate novel activates for specific targets based on users' input. The ChemGenerator relies on an autoencoder-based algorithm of Recurrent Neural Networks with Long Short-Term Memory by training of 7 million of molecular Simplified Molecular-Input Line-Entry System as the basic model, and further develops target guided generation by transfer learning. As results, ChemGenerator gains lower loss (<0.01) than existing reference model (0.2~0.4) and shows good performance in the case of Epidermal Growth Factor Receptor. Meanwhile, ChemGenerator is now freely accessible to the public by http://smiles.tcmobile.org. In proportion to endless molecular enumeration and time-consuming expensive experiments, this work demonstrates an efficient alternative way for the first virtual screening in drug discovery.

A natural language processing approach based on embedding deep learning from heterogeneous compounds for quantitative structure-activity relationship modeling.

Over the past decade, rapid development in biological and chemical technologies such as high-throughput screening, parallel synthesis, has been significantly increased the amount of data, which requires the creation and the integration of new analytical methods, especially deep learning models. Recently, there is an increasing interest in deep learning utilization in computer-aided drug discovery due to its exceptional successful application in many fields. The present work proposed a natural language processing approach, based on embedding deep neural networks. Our method aims to transform the Simplified Molecular Input Line Entry System format into word embedding vectors to represent the semantics of compounds. These vectors are fed into supervised machine learning algorithms such as convolutional long short-term memory neural network, support vector machine, and random forest to build up quantitative structure-activity relationship models on toxicity data sets. The obtained results on toxicity data to the ciliate Tetrahymena pyriformis (IGC50 ), and acute toxicity rat data expressed as median lethal dose of treated rats (LD50 ) show that our approach can eventually be used to predict the activities of chemical compounds efficiently. All material used in this study is available online through the GitHub portal (https://github.com/BoukeliaAbdelbasset/NLPDeepQSAR.git).

QSAR modeling of toxicities of ionic liquids toward Staphylococcus aureus using SMILES and graph invariants

Ionic liquids (ILs) have been popular in many industrial and chemical processes, like antimicrobial properties, solvents, and synthesis of new compounds with antioxidant activity. Because of the significance of their application, the prediction minimal inhibitory concentration (MIC) of 204 ILs and the minimal bactericidal concentration (MBC) of 114 ILs of them against Staphylococcus aureus (S. aureus) have been carried out using the quantitative structure activity relationship (QSAR) based on the Monte Carlo method. Using the simplified molecular input line entry system (SMILES) notation, molecular structures of all of ILs were displayed. Hybrid optimal descriptor was employed in developing the model for pMIC and pMBC, which was obtained by combining the molecular graph and SMILES. For pMIC, hybrid optimal descriptors were calculated via SMILES and hydrogen-suppressed molecular graph (HSG), as well as hybrid optimal descriptors for pMBC were calculated via SMILES and hydrogen-filled graph (HFG). The total dataset was randomly split into training, invisible training, calibration, and validation set for three times. Statistically analyzed by the calculated descriptors, a QSAR model was developed for pMIC and pMBC of ILs, and the index of ideality of correlation (IIC) was examined as a benchmark for predictive potential of these models. Their correlation coefficient (R2) values of the training, invisible training, calibration, and validation sets for three splits were 0.8585–0.8853, 0.8523–0.8898, 0.8809–0.9240, and 0.8036–0.8903 for pMIC and 0.8357–0.8991, 0.8223–0.9306, 0.8372–0.9170, and 0.8171–0.8901 for pMBC, respectively. The results show that the predictability to develop the QSAR model for all splits is at a high level.

In-silico identification of fingerprint of pyrazolyl sulfonamide responsible for inhibition of N-myristoyltransferase using Monte Carlo method with index of ideality of correlation

Human African trypanosomiasis (HAT) or sleeping sickness like infections remain a serious health concern around the globe due to unavailability of safe and potential drugs for their treatment. Moreover, developing safe, potential and highly specific target based treatments is still a challenge for present drug discovery programs. A series of pyrazole based sulfonamides are identified as an inhibitor of Trypanosoma brucei N-myristoyltransferase (TbNMT). In the present manuscript, we have developed robust and reliable QSAR models by using the balance of correlation method in CORAL software. The chemical structures are represented by simplified molecular input line entry system (SMILES). The significance of the index of ideality correlation (IIC) with applicability domain (AD) is also studied at depth. The models developed by considering the index of ideality of correlation (IIC) were found to statistically more significant and robust. One QSAR model with best = 0.8638 for split 2 was considered as the leading model. A greater value of cRp2 i.e. 0.5 for all models in Y-randomization test showed the robustness of developed models. The outliers and promoters of increase and decrease of endpoint were also extracted independently from the leading models. The mechanistic interpretation of developed models explains the role of different structural attributes in predicting the pIC50 of pyrazole sulfonamides extracted from the crystal structure of Leishmania major N-myristoyltransferase (NMT) along with co-crystallized myristoyl-CoA and ligands NMT106, NMT157, NMT187 and NMT236 (PDB ID: 4A2Z, 4A30, 4A32, 2WSA). Communicated by Ramaswamy H. Sarma

MBLinhibitors.com, a Website Resource Offering Information and Expertise for the Continued Development of Metallo--Lactamase Inhibitors.

In an effort to facilitate the discovery of new, improved inhibitors of the metallo-β-lactamases (MBLs), a new, interactive website called MBLinhibitors.com was developed. Despite considerable efforts from the science community, there are no clinical inhibitors of the MBLs, which are now produced by human pathogens. The website, MBLinhibitors.com, contains a searchable database of known MBL inhibitors, and inhibitors can be searched by chemical name, chemical formula, chemical structure, Simplified Molecular-Input Line-Entry System (SMILES) format, and by the MBL on which studies were conducted. The site will also highlight a “MBL Inhibitor of the Month”, and researchers are invited to submit compounds for this feature. Importantly, MBLinhibitors.com was designed to encourage collaboration, and researchers are invited to submit their new compounds, using the “Submit” function on the site, as well as their expertise using the “Collaboration” function. The intention is for this site to be interactive, and the site will be improved in the future as researchers use the site and suggest improvements. It is hoped that MBLinhibitors.com will serve as the one-stop site for any important information on MBL inhibitors and will aid in the discovery of a clinically useful MBL inhibitor.

Bidirectional Molecule Generation with Recurrent Neural Networks.

Recurrent neural networks (RNNs) are able to generate de novo molecular designs using simplified molecular input line entry systems (SMILES) string representations of the chemical structure. RNN-based structure generation is usually performed unidirectionally, by growing SMILES strings from left to right. However, there is no natural start or end of a small molecule, and SMILES strings are intrinsically nonunivocal representations of molecular graphs. These properties motivate bidirectional structure generation. Here, bidirectional generative RNNs for SMILES-based molecule design are introduced. To this end, two established bidirectional methods were implemented, and a new method for SMILES string generation and data augmentation is introduced-the bidirectional molecule design by alternate learning (BIMODAL). These three bidirectional strategies were compared to the unidirectional forward RNN approach for SMILES string generation, in terms of the (i) novelty, (ii) scaffold diversity, and (iii) chemical-biological relevance of the computer-generated molecules. The results positively advocate bidirectional strategies for SMILES-based molecular de novo design, with BIMODAL showing superior results to the unidirectional forward RNN for most of the criteria in the tested conditions. The code of the methods and the pretrained models can be found at URL https://github.com/ETHmodlab/BIMODAL.

Prediction of Oral Bioavailability in Rats: Transferring Insights from in Vitro Correlations to (Deep) Machine Learning Models Using in Silico Model Outputs and Chemical Structure Parameters.

Oral administration of drug products is a strict requirement in many medical indications. Therefore, bioavailability prediction models are of high importance for prioritization of compound candidates in the drug discovery process. However, oral exposure and bioavailability are difficult to predict, as they are the result of various highly complex factors and/or processes influenced by the physicochemical properties of a compound, such as solubility, lipophilicity, or charge state, as well as by interactions with the organism, for instance, metabolism or membrane permeation. In this study, we assess whether it is possible to predict intravenous (iv) or oral drug exposure and oral bioavailability in rats. As input parameters, we use (i) six experimentally determined in vitro and physicochemical endpoints, namely, membrane permeation, free fraction, metabolic stability, solubility, pKa value, and lipophilicity; (ii) the outputs of six in silico absorption, distribution, metabolism, and excretion models trained on the same endpoints, or (iii) the chemical structure encoded as fingerprints or simplified molecular input line entry system strings. The underlying data set for the models is an unprecedented collection of almost 1900 data points with high-quality in vivo experiments performed in rats. We find that drug exposure after iv administration can be predicted similarly well using hybrid models with in vitro- or in silico-predicted endpoints as inputs, with fold change errors (FCE) of 2.28 and 2.08, respectively. The FCEs for exposure after oral administration are higher, and here, the prediction from in vitro inputs performs significantly better in comparison to in silico-based models with FCEs of 3.49 and 2.40, respectively, most probably reflecting the higher complexity of oral bioavailability. Simplifying the prediction task to a binary alert for low oral bioavailability, based only on chemical structure, we achieve accuracy and precision close to 70%.

Investigation of Machine Intelligence in Compound Cell Activity Classification.

Machine intelligence has been greatly developed in the past decades and has been widely used in many fields. In the recent years, many reports have shown its satisfactory effect in drug discovery. In this study, machine intelligence methods were explored to assist the cell activity prediction. Multiple machine intelligence methods including support vector machine, decision tree, random forest, extra trees, gradient boosting machine, convolutional neural network, long short-term memory network, and gated recurrent unit network were employed to separate compounds based on their cell activity. Different from some reported classification models, compounds were expressed as a string by the simplified molecular input line entry system and directly used as input rather than any chemical descriptors, which mimicked natural language processing. Both the single cell strain and whole data set under the balanced and imbalanced data distributions were discussed, respectively. Different activity cutoffs were set for the single (Z-score = 3) and the whole (Z-score = 5 and 6) data set. Nine metrics were used to evaluate the models including accuracy, precision, recall, f1-score, area under the receiver operating characteristic curve score, Cohen's κ, Brier score, Matthews correlation coefficient, and balanced accuracy. The results show that the gradient boosting machine is competent at balanced data distribution, and convolutional neural network is qualified for the imbalanced one. The results demonstrate that both classic machine learning methods and deep learning methods have potential in classification of compound cell activity.

QSPR Modeling For Critical Temperatures Of Organic Compounds Using Hybrid Optimal Descriptors

The simplified molecular input line entry system (SMILES) is particularly suitable for high-speed machine processing, based on the Monte Carlo method using CORAL software. Quantitative structure-property relationships (QSPR) of critical temperatures have been established using a dataset of 165 diverse organic compounds employing hybrid optimal descriptors defined by graph and SMILES notation. External validation is one of the most important parts in the evaluation of model performance. However, previous models on the same dataset have poor predictive power in the external test set, or the authors had not done that check. In the present work, the predictive ability of model has been tested using external validation. The statistical quality of the three splits are similar and good. The r2 values for the best model are: r2 = 0.98 for the training set, r2 = 0.95 for the calibration set, and r2 = 0.94 for the validation set.

Improving prediction of phenotypic drug response on cancer cell lines using deep convolutional network

BackgroundUnderstanding the phenotypic drug response on cancer cell lines plays a vital role in anti-cancer drug discovery and re-purposing. The Genomics of Drug Sensitivity in Cancer (GDSC) database provides open data for researchers in phenotypic screening to build and test their models. Previously, most research in these areas starts from the molecular fingerprints or physiochemical features of drugs, instead of their structures.ResultsIn this paper, a model called twin Convolutional Neural Network for drugs in SMILES format (tCNNS) is introduced for phenotypic screening. tCNNS uses a convolutional network to extract features for drugs from their simplified molecular input line entry specification (SMILES) format and uses another convolutional network to extract features for cancer cell lines from the genetic feature vectors respectively. After that, a fully connected network is used to predict the interaction between the drugs and the cancer cell lines. When the training set and the testing set are divided based on the interaction pairs between drugs and cell lines, tCNNS achieves 0.826, 0.831 for the mean and top quartile of the coefficient of determinant (R2) respectively and 0.909, 0.912 for the mean and top quartile of the Pearson correlation (Rp) respectively, which are significantly better than those of the previous works (Ammad-Ud-Din et al., J Chem Inf Model 54:2347–9, 2014), (Haider et al., PLoS ONE 10:0144490, 2015), (Menden et al., PLoS ONE 8:61318, 2013). However, when the training set and the testing set are divided exclusively based on drugs or cell lines, the performance of tCNNS decreases significantly and Rp and R2 drop to barely above 0.ConclusionsOur approach is able to predict the drug effects on cancer cell lines with high accuracy, and its performance remains stable with less but high-quality data, and with fewer features for the cancer cell lines. tCNNS can also solve the problem of outliers in other feature space. Besides achieving high scores in these statistical metrics, tCNNS also provides some insights into the phenotypic screening. However, the performance of tCNNS drops in the blind test.

Predicting Value of Binding Constants of Organic Ligands to Beta-Cyclodextrin: Application of MARSplines and Descriptors Encoded in SMILES String

The quantitative structure–activity relationship (QSPR) model was formulated to quantify values of the binding constant (lnK) of a series of ligands to beta–cyclodextrin (β-CD). For this purpose, the multivariate adaptive regression splines (MARSplines) methodology was adopted with molecular descriptors derived from the simplified molecular input line entry specification (SMILES) strings. This approach allows discovery of regression equations consisting of new non-linear components (basis functions) being combinations of molecular descriptors. The model was subjected to the standard internal and external validation procedures, which indicated its high predictive power. The appearance of polarity-related descriptors, such as XlogP, confirms the hydrophobic nature of the cyclodextrin cavity. The model can be used for predicting the affinity of new ligands to β-CD. However, a non-standard application was also proposed for classification into Biopharmaceutical Classification System (BCS) drug types. It was found that a single parameter, which is the estimated value of lnK, is sufficient to distinguish highly permeable drugs (BCS class I and II) from low permeable ones (BCS class II and IV). In general, it was found that drugs of the former group exhibit higher affinity to β-CD then the latter group (class III and IV).

Drug Analogs from Fragment-Based Long Short-Term Memory Generative Neural Networks.

Several recent reports have shown that long short-term memory generative neural networks (LSTM) of the type used for grammar learning efficiently learn to write Simplified Molecular Input Line Entry System (SMILES) of druglike compounds when trained with SMILES from a database of bioactive compounds such as ChEMBL and can later produce focused sets upon transfer learning with compounds of specific bioactivity profiles. Here we trained an LSTM using molecules taken either from ChEMBL, DrugBank, commercially available fragments, or from FDB-17 (a database of fragments up to 17 atoms) and performed transfer learning to a single known drug to obtain new analogs of this drug. We found that this approach readily generates hundreds of relevant and diverse new drug analogs and works best with training sets of around 40,000 compounds as simple as commercial fragments. These data suggest that fragment-based LSTM offer a promising method for new molecule generation.

Development of prediction model for fructose- 1,6- bisphosphatase inhibitors using the Monte Carlo method

ABSTRACTFructose-1,6-bisphosphatase (FBPase) is an enzyme important for regulation of gluconeogenesis, which is a major process in the liver responsible for glucose production. Inhibition of FBPase enzyme causing blockage of the gluconeogenesis process represents a newer scheme in the progress of anti-diabetic drugs. The current research describes the development of hybrid optimal descriptors-based quantitative structure–activity relationship (QSAR) models intended for a set of 62 FBPase inhibitors with the Monte Carlo method. The molecular structures were expressed by the simplified molecular input line entry system (SMILES) notation. Three splits were prepared by random division of the molecules into training set, calibration set and validation set. Statistical parameters obtained from QSAR modelling were good for various designed splits. The best QSAR model showed the following parameters: the values of r2 for calibration set and validation set of the best model were 0.6837 and 0.8623 and of Q2 were 0.6114 and 0.8036, respectively. Based on the results obtained for correlation weights, different structural attributes were described as promoter of the endpoint. Further, these structural attributes were used in designing of new FBPase inhibitors and a molecular docking study was completed for the determination of interactions of the designed molecules with the enzyme.

AutophagySMDB: a curated database of small molecules that modulate protein targets regulating autophagy

ABSTRACTMacroautophagy/autophagy is a complex self-degradative mechanism responsible for clearance of non functional organelles and proteins. A range of factors influences the autophagic process, and disruptions in autophagy-related mechanisms lead to disease states, and further exacerbation of disease. Despite in-depth research into autophagy and its role in pathophysiological processes, the resources available to use it for therapeutic purposes are currently lacking. Herein we report the Autophagy Small Molecule Database (AutophagySMDB; http://www.autophagysmdb.org/) of small molecules and their cognate protein targets that modulate autophagy. Presently, AutophagySMDB enlists ~10,000 small molecules which regulate 71 target proteins. All entries are comprised of information such as EC50 (half maximal effective concentration), IC50 (half maximal inhibitory concentration), Kd (dissociation constant) and Ki (inhibition constant), IUPAC name, canonical SMILE, structure, molecular weight, QSAR (quantitative structure activity relationship) properties such as hydrogen donor and acceptor count, aromatic rings and XlogP. AutophagySMDB is an exhaustive, cross-platform, manually curated database, where either the cognate targets for small molecule or small molecules for a target can be searched. This database is provided with different search options including text search, advanced search and structure search. Various computational tools such as tree tool, cataloging tools, and clustering tools have also been implemented for advanced analysis. Data and the tools provided in this database helps to identify common or unique scaffolds for designing novel drugs or to improve the existing ones for autophagy small molecule therapeutics. The approach to multitarget drug discovery by identifying common scaffolds has been illustrated with experimental validation.Abbreviations: AMPK: AMP-activated protein kinase; ATG: autophagy related; AutophagySMDB: autophagy small molecule database; BCL2: BCL2, apoptosis regulator; BECN1: beclin 1; CAPN: calpain; MTOR: mechanistic target of rapamycin kinase; PPARG: peroxisome proliferator activated receptor gamma; SMILES: simplified molecular input line entry system; SQSTM1: sequestosome 1; STAT3: signal transducer and activator of transcription

Design and development of novel focal adhesion kinase (FAK) inhibitors using Monte Carlo method with index of ideality of correlation to validate QSAR

ABSTRACTQuantitative structure-activity relationship (QSAR) modelling of 55 focal adhesion kinase (FAK) (EC 2.7.10.2) inhibitors of triazinic nature was performed using the Monte Carlo method. The QSAR models were designed by CORAL software, and optimal descriptors were calculated with the simplified molecular input line entry system (SMILES). Four splits were made from the triazinic derivative data by random division into training, invisible training, calibration and validation sets. The QSAR results from these four random splits were robust, very simple, predictive and reliable. The best statistical parameters of the validation set (r2 = 0.8398 and Q2 = 0.7722) for the QSAR equation for split 3 with IIC = 0.9127 were obtained. The predictive potential of QSAR models of FAK inhibitors was explored by applying the index of ideality of correlation (IIC), which is a new criterion for the prediction of the potential for quantitative structure–property activity relationships (QSPRs/QSARs). The present method follows OECD principles.

Prediction of the adsorption coefficients of some aromatic compounds on multi-wall carbon nanotubes by the Monte Carlo method

ABSTRACTIn this investigation, quantitative structure–property relationship (QSPR) modelling of adsorption coefficients of 69 aromatic compounds on multi-wall carbon nanotubes (MWCNTs) was studied using the Monte Carlo method. QSPR models were calculated with CORAL software, and optimal descriptors were calculated with the simplified molecular input line entry system (SMILES) and hydrogen-suppressed molecular graphs (HSGs). The aromatic compound data set was randomly split into training, invisible training, calibration and validation sets. Analysis of three probes of the Monte Carlo optimization with three random splits was done. The results from three random splits displayed robust, very simple, predictable and reliable models for the training, invisible training, calibration and validation sets with a coefficient of determination (r2) equal to 0.9463–0.8528, 0.9020–0.8324, 0.9606–0.9178 and 0.9573–0.8228, respectively. As a result, the models obtained help to identify the hybrid descriptors for the increase and the decrease of the adsorption coefficient of aromatic compounds on MWCNTs. This simple QSPR model can be used for the prediction of the adsorption coefficient of numerous aromatic compounds on MWCNTs.

End-to-End Representation Learning for Chemical-Chemical Interaction Prediction.

Chemical-chemical interaction (CCI) plays a major role in predicting candidate drugs, toxicities, therapeutic effects, and biological functions. CCI is typically inferred from a variety of information; however, CCI has yet not been predicted using a learning-based approach. In other drug analyses, deep learning has been actively used in recent years. However, in most cases, deep learning has been used only for classification even though it has feature extraction capabilities. Thus, in this paper, we propose an end-to-end representation learning method for CCI, named DeepCCI, which includes feature extraction and a learning-based approach. Our proposed architecture is based on the Siamese network. Hidden representations are extracted from a simplified molecular input line entry system (SMILES), which is a string notation representing the chemical structure using weight-shared convolutional neural networks. Subsequently, L1 element-wise distances between the two extracted hidden representations are measured. The performance of DeepCCI is compared with those of 12 fingerprint-method combinations. The proposed DeepCCI shows the best performance in most of the evaluation metrics used. In addition, DeepCCI was experimentally validated to guarantee the commutative property. The automatically extracted features can alleviate the efforts required for manual feature engineering and improve prediction performance.

A novel methodology on distributed representations of proteins using their interacting ligands.

MotivationThe effective representation of proteins is a crucial task that directly affects the performance of many bioinformatics problems. Related proteins usually bind to similar ligands. Chemical characteristics of ligands are known to capture the functional and mechanistic properties of proteins suggesting that a ligand-based approach can be utilized in protein representation. In this study, we propose SMILESVec, a Simplified molecular input line entry system (SMILES)-based method to represent ligands and a novel method to compute similarity of proteins by describing them based on their ligands. The proteins are defined utilizing the word-embeddings of the SMILES strings of their ligands. The performance of the proposed protein description method is evaluated in protein clustering task using TransClust and MCL algorithms. Two other protein representation methods that utilize protein sequence, Basic local alignment tool and ProtVec, and two compound fingerprint-based protein representation methods are compared.ResultsWe showed that ligand-based protein representation, which uses only SMILES strings of the ligands that proteins bind to, performs as well as protein sequence-based representation methods in protein clustering. The results suggest that ligand-based protein description can be an alternative to the traditional sequence or structure-based representation of proteins and this novel approach can be applied to different bioinformatics problems such as prediction of new protein–ligand interactions and protein function annotation.Availability and implementation https://github.com/hkmztrk/SMILESVecProteinRepresentation Supplementary information Supplementary data are available at Bioinformatics online.

Quantitative Structure-activity Relationship Study of Betulinic Acid Derivatives Against HIV using SMILES-based Descriptors.

Human Immunodeficiency Virus (HIV) is the causative agent of Acquired Immunodeficiency Syndrome (AIDS) that imposes a global health burden. Therefore, HIV therapeutic agents have been discovery and development. To construct Quantitative-structure Activity Relationship (QSAR) models of betulinic acid derivatives with anti-HIV activity using Simplified Molecular-Input Line-Entry System (SMILES)- based descriptors. A data set of 107 betulinic acid derivatives and their anti-HIV activity was used to develop QSAR models. The SMILES format of the compounds was employed as descriptors for model construction using the CORAL software by means of the Monte Carlo method. Constructed QSAR models provided good correlation coefficients (R2) and root mean square error (RMSE) with values in the range of 0.5660-0.5890 and 0.963-1.020, respectively, for the training set, R2 value of 0.7206-0.7837 and RMSE as 0.609-1.250, respectively, for the calibration set, and R2 value of 0.6257-0.7748 and RMSE as 0.837-0.995, respectively, for the validation set. The best QSAR model displayed statistical parameters for training set: R2 = 0.5660 and RMSE = 0.963; calibration set: R2 = 0.7273 and RMSE = 0.609, and validation set: R2 = 0.7748 and RMSE = 0.972. In addition, features of the molecular structure that are promoters of the endpoint increase and decrease were defined and discussed. These are the basis for the mechanistic interpretation of the suggested models. These findings provide useful knowledge for guiding the design of novel compounds with promising anti-HIV activity.

Towards the Development of Global Nano-Quantitative Structure-Property Relationship Models: Zeta Potentials of Metal Oxide Nanoparticles.

Zeta potential indirectly reflects a charge of the surface of nanoparticles in solutions and could be used to represent the stability of the colloidal solution. As processes of synthesis, testing and evaluation of new nanomaterials are expensive and time-consuming, so it would be helpful to estimate an approximate range of properties for untested nanomaterials using computational modeling. We collected the largest dataset of zeta potential measurements of bare metal oxide nanoparticles in water (87 data points). The dataset was used to develop quantitative structure–property relationship (QSPR) models. Essential features of nanoparticles were represented using a modified simplified molecular input line entry system (SMILES). SMILES strings reflected the size-dependent behavior of zeta potentials, as the considered quasi-SMILES modification included information about both chemical composition and the size of the nanoparticles. Three mathematical models were generated using the Monte Carlo method, and their statistical quality was evaluated (R2 for the training set varied from 0.71 to 0.87; for the validation set, from 0.67 to 0.82; root mean square errors for both training and validation sets ranged from 11.3 to 17.2 mV). The developed models were analyzed and linked to aggregation effects in aqueous solutions.