Quantitative Structure-property Relationship Models Research Articles

Transformation of phenol and nitrobenzene by superoxide radicals: Kinetics and mechanisms

In this study, the kinetics and mechanisms of superoxide radical (O2·-)-mediated transformation of two monosubstituted aromatic contaminants with diametrically opposed electronic effect, namely phenol (PhOH) and nitrobenzene (NB), were investigated through combined experimental and theoretical approaches. The reaction rate constant between O2·- and PhOH was determined to be 269 M−1 s−1via competition kinetic method. Such weak role of O2·- was confirmed by density functional theory (DFT) calculations at SMD/CAM-B3LYP/6-311++G(2d,2p)//CAM-B3LYP/6-31+G(d,p) level of theory. However, NB exhibited a higher reactivity towards O2·- than PhOH, demonstrating the selectivity of O2·- to electron withdrawing group. Despite the difference in reactivity, O2·- shows very limited removal capacity for PhOH and NB. These results were also in reasonably good agreement with an indication from a quantitative structure–property relationship model. With these quantitative and mechanistic results, we provided a more comprehensive understanding on the O2·--mediated transformation of organic pollutants bearing different electron donating/withdrawing groups.

A simple correlation for reliable prediction of intrinsic viscosity (limiting viscosity number) of different polymer-solvent combinations

Intrinsic viscosity (limiting viscosity number) of a dilute polymer solution is one of the most fundamental properties of polymers for solubility parameters calculation, determining the molecular weight, size, and topological structure of polymers, and other physicochemical property characterization. The purpose of this work is to introduce a core correlation for predicting intrinsic viscosity of different polymer-solvent combinations from only the structure of the repeating unit structure of polymer and solubility parameter of polymer. Model's reliability can be enhanced by considering two correcting functions based on the effective dispersion, polar, and hydrogen bonding components of solubility for some solvents. The largest available experimental data including intrinsic viscosity of 74 polymer-solvent combinations are used to derive and test the improved model. The predicted intrinsic viscosity of the new model show higher reliability as compared with the results of one empirical and two quantitative structure-property relationships (QSPR) approaches for 65 polymer-solvent combinations. The value of the Root Mean Square Error (RMSE) of the improved model is 24.05 cm3/g, which is less than three comparative models, i.e. 86.08, 37.79, and 34.63 cm3/g. The new method also gives good results as compared to two further complex QSPR models for nine extra polymer-solvent combinations. Moreover, different statistical parameters confirm excellent reliability of the improved correlation as compared to the best accessible predictive methods.

An accurate and interpretable deep learning model for environmental properties prediction using hybrid molecular representations

AbstractLipophilicity, as quantified by the decimal logarithm of the octanol–water partition coefficient (logKOW), is an essential environmental property. Deep neural networks (DNNs) based quantitative structure–property relationship (QSPR) studies have received more and more attention because of their excellent performance for prediction. However, the black‐box nature of DNNs limits the application range where interpretability is essential. Hence, this study aims to develop an accurate and interpretable deep neural network (AI‐DNN) model for logKOWprediction. A hybrid method of molecular representation was employed to guarantee the accuracy of the proposed AI‐DNN model. The hybrid molecular representations are able to integrate the directed message passing neural networks (D‐MPNNs) learned molecular representations and the fixed molecule‐level features of CDK descriptors, and can capture both the local and the global features of overall molecule. The performance analysis shows that the proposed QSPR model exhibits promising predictive accuracy and discriminative power in the structural isomers and stereoisomers. Moreover, the Monte Carlo Tree Search (MCTS) approach was used to interpret the proposed AI‐DNN model by identifying the molecular substructures contributed to the lipophilicity. This interpretability can be applied to critical fields where there is a high demand for interpretable deep networks, such as green solvent design and drug discovery.

Turning deep-eutectic solvents into value-added products for CO2 capture: A desirability-based virtual screening study

Carbon dioxide capture and storage is one of the primary mitigation approaches endorsed by current global climate change policies, since it is the largest contributing gas to the greenhouse effect. Recently, deep eutectic solvents (DESs) have been found as promising green solvents for CO2 capture, particularly in industrial applications. Compared to conventional ionic liquids, for example, DESs display similar solvent properties but they are far cheaper, easier to prepare, and greener. Yet their major drawback is linked to their high viscosity that impedes management not fulfilling industrial demands. Further efforts are thus required to develop new DES solvents that hold together high capacity for CO2 uptake and low viscosity. In this work, we adopted a desirability-based decision approach to jointly handle these two conflicting properties. We began by setting up Quantitative Structure-Property Relationships (QSPR) models for uncovering the two properties based on a dataset of known binary DESs. Then, desirability functions derived from the individual QSPR models were found and combined into an overall desirability-based criterion that was applied to screen and rank two libraries, one comprising experimentally reported DESs and another one with newly designed DESs. The latter enabled us to propose novel efficient DESs for CO2 uptake, i.e., with the most suitable trade-offs between CO2 absorption capacity and viscosity. Finally, but most importantly, this work demonstrates the usefulness of the desirability-based approach for the rational discovery of deep eutectic solvents or other materials to suit particular sustainable applications.

Synergistic Behavior of Plant Proteins and Biobased Latexes in Bioplastic Food Packaging Materials: Experimental and Machine Learning Study.

Plant-based proteins are attractive components which may serve as sustainable alternatives to current petrochemical products. Both soy protein and major corn protein, zein, are of interest in food packaging applications due to their sustainability, biodegradation properties, and inherent physicochemical properties. This study discusses the development of bioplastic materials, where it explores the effects of combining zein, soy protein, and plasticizing latexes derived from plant oil-based monomers (POBMs) on properties of resulting bioplastic films. By looking for synergistic effects of soy protein's inherent film formation ability and zein's higher strength, we prepare strong yet flexible soy-zein films as materials, called proteoposites. Incorporation of natural additive POBM-latexes helps to plasticize and hydrophobize the bioplastic films and thus to improve mechanical and barrier properties. Variation of the POBM-latexes' particle size further aims to enhance the performance of resulting bioplastic films. As a result, modified soy-zein proteoposite films with improved moisture resistance, enhanced mechanical behavior, and greater barrier properties were developed. Machine learning-based computational models were utilized in order to find main structural factors affecting the bioplastic's properties and develop a quantitative structure-property relationship model between the physicochemical properties of the film components and the resulted bioplastics' properties and performance. The developed model effectively predicts experimental outcomes with >85% (R2: 0.85) accuracy. The newly synthesized proteoposites confirmed the machine learning model predictions. As a result, proteoposite films made of two plant proteins and modified with POBM-latexes can be considered as an attractive and viable replacement for petrochemical food packaging products.

A structure-property model for the prediction of pKa values of different amines in the CO2 capture process of concern to the prediction of thermodynamic properties

In this study, Quantitative Structure-Property Relationship (QSPR) method has been applied on a dataset including the pKa values of 21 different structures (20 amines and ammonia) at four temperatures in the post-combustion-capture process using QSARINS. QSPR models have been accomplished in two situations, called fixed and unfixed temperatures. In the fixed temperatures, four predictive models with one descriptor have been separately developed at each temperature. Besides, thermodynamic properties have been calculated by the van’t Hoff equation using predicted pKa values. In the unfixed temperatures, two general models were developed which take into account the effects of molecular structure and temperature, simultaneously. The predicted pKa values and calculated Gibbs free energy have a good consistency with related values in the literature. The reliability of each developed model has been examined using external assessments. The values of some statistical parameters such as R2 = 0.89 and AARD% = 2.68 verify the prediction capability of models.

Threshold Sooting Index of Sustainable Aviation Fuel Candidates from Composition Input Alone: Progress toward Uncertainty Quantification

A quantitative structure–property relationship model has been developed to predict the threshold sooting index (TSI) of arbitrary mixtures of aliphatic and aromatic hydrocarbons of known composition. The model employs contributions from eight molecular fragments plus a global shift and a penalty factor for naphthenic compounds. For each coefficient, the contributions were determined by a constrained regression to data from five different experimental campaigns, which were stitched together by setting the TSI of methylcyclohexane to 5 and the TSI of 1-methylnaphthalene to 100. Unique to this study; the TSI of 1,3,5-trimethylbenzene was restricted to the range of 54.7 and 63.1, which significantly constrains uncertainty. Within the composite training dataset, which contained 65 molecules and 124 data points, including simple mixtures, the model was found to match 95% of the data within 8.9 TSI. Validation of the model against n-butylcyclohexane, dimethylcyclooctane, and a six-component surrogate jet fuel shows prediction to be well within the 95-percentile confidence band of the experiment. This model is the first to integrate the linear blending rule for the TSI with a linear quantitative structure–property relationship model for the TSI, and the first time that referee controls have been applied to ensure that all datasets, experimental and modeled, are normalized to the same scale.

Benchmarking machine learning methods for modeling physical properties of ionic liquids

The great importance of the ability to quantitatively predict the properties of ionic liquids (ILs) using quantitative structure–property relationships (QSPR) models necessitates the understanding of which modern machine learning (ML) methods in combination with which types of molecular representations are preferable to use for this purpose. To address this problem, a large-scale benchmarking study of QSPR models built by combining three traditional ML methods and neural networks with seven different architectures with five types of molecular representations (in the form of either numerical molecular descriptors or SMILES text strings) to predict six important physical properties of ILs (density, electrical conductance, melting point, refractive index, surface tension, and viscosity) was carried out. The datasets include from 407 to 1204 diverse ILs composed of various organic and inorganic ions. QSPR models for predicting the properties of ILs at eight different temperatures were built using multi-task learning. The best combinations of ML methods and molecular representations were identified for each of the properties. A unified ranking system was introduced to rank and prioritize different ML methods and molecular representations. It was shown in this study that on average: (i) nonlinear ML methods perform much better than linear ones, (ii) neural networks perform better than traditional ML methods, (iii) Transformers, which are actively used in natural language processing (NLP), perform better than other types of neural networks due to the advanced ability to analyze chemical structures of ILs encoded into SMILES text strings. A special “component-wise” cross-validation scheme was applied to assess how much the predictive performance deteriorates for the ILs composed of cations and anions that are not present in the dataset.

Technique of Augmenting Molecular Graph Data by Perturbating Hidden Features.

Quantitative structure-property relationship models are useful in efficiently searching for molecules with desired properties in drug discovery and materials development. In recent years, many such models based on graph neural networks, showing good prediction performance, have been reported. Training graph neural networks generally require many samples, but by using a training method for a small dataset, it is possible to extract features that enable successful prediction. Herein, we design a method of augmenting graph data. In this method, random perturbations are added with a certain probability to some vertex features during message passing. We verify the proposed method's effectiveness in regression and classification tasks. It is confirmed that the proposed method is effective when the perturbation is added immediately before the readout of the graph neural network, and the effect of the data augmentation is most evident for small datasets of approximately 1000 samples.

Prediction of Carbonate Selectivity of PVC-Plasticized Sensor Membranes with Newly Synthesized Ionophores through QSPR Modeling

Developing a potentiometric sensor with required target properties is a challenging task. This work explores the potential of quantitative structure-property relationship (QSPR) modeling in the prediction of potentiometric selectivity for plasticized polymeric membrane sensors based on newly synthesized ligands. As a case study, we have addressed sensors with selectivity towards carbonate—an important topic for environmental and biomedical studies. Using the logKsel(HCO3−/Cl−) selectivity data on 40 ionophores available in literature and their substructural molecular fragments as descriptors, we have constructed a QSPR model, which has demonstrated reasonable precision in predicting selectivities for newly synthesized ligands sharing similar molecular fragments with those employed for modeling.

User-assisted methodology targeted for building structure interpretable QSPR models for boosting CO2 capture with ionic liquids

Task-specific ionic liquid (IL) is an emerging class of compounds that may be environmentally friendly. Properly selected, these compounds may be green alternative to amine solutions and can replace them in post-combustion carbon dioxide (CO2) capture processes on an industrial scale. However, owing to the vast diversity of ions and their possible combinations, laboratory research is time consuming and expensive. Therefore, computational methods are preferred for assessing their potential applications. In this study, three quantitative structure–property relationship models based on six distinct descriptors were created to predict Henry’s law constant (HLC) of CO2 in 62 ILs. The statistical parameters of multiple linear regression, logistic regression, and partial least squares models were satisfactory. In all cases, the coefficients of determination (R2) exceeded 0,90, and both external and internal validation proved them to be reliable and predictive with Q2 and Rpred2 values exceeding 0,90 and 0,87, respectively. Three of the descriptors were attributed to cations, and three were attributed to anions. In contrast to many other models, the descriptors were chosen in a manner that ensured their interpretability. Each of the six descriptors was analyzed for its influence on HLC. On this basis, guidelines for designing the structure of ionic liquids with increased CO2 absorption capacity were developed.

Quantitative structure-property relationship for predicting the diffusion coefficient of ionic liquids

Quantitative structure–property relationship (QSPR) models can offer theoretical prediction of physicochemical properties of ionic liquids (ILs), and the exploration of molecular descriptors that reflect the structural characteristics of ILs is of primary importance for QSPR models. Here, three novel QSPR models, termed as Model I, Model II and DB∞-QSPR model, are developed to predict the diffusion coefficient of ILs by using conductor-like screening model for segment activity coefficient (COSMO-SAC) based descriptors. The established Model I matches well with the experimental data set for both cations and anions, as evidenced by the high values of the coefficient of determination (R2) (greater than 0.95) in the training set and the testing set. Model II can be used to calculate the self-diffusion coefficients of ILs with a fixed anion of [NTf2], and the corresponding R2 values are greater than 0.97 in the training set and the testing set. The mutual diffusion coefficients of ILs in water at infinite dilution (DB∞) can reliably be predicted by DB∞-QSPR model, which demonstrates excellent predictive power because of the high values of R2 (greater than 0.97) in the training set and the testing set. In addition, the value of average absolute relative deviation (AARD) between experimental and calculated DB∞ is 5.023% in the overall set. According to the afore-mentioned QSPR models, the diffusion coefficient of ILs is positively correlated with temperature and negatively correlated with the cavity volume of ion.

Data-driven selection of electrolyte additives for aqueous magnesium batteries

Data-driven quantitative structure–property relationship models facilitate the selection of potent electrolyte additives for aqueous magnesium batteries in an active design of experiments approach.

Probabilistic Mean Quantitative Structure–Property Relationship Modeling of Jet Fuel Properties

We present a novel probabilistic mean quantitative structure–property relationship (M-QSPR) method for the prediction of jet fuel properties considering two-dimensional gas chromatography measurements. Fuels are represented as one mean pseudo-structure that is inferred by a weighted average over structures of 1866 molecules that could be present in the individual fuel. The method allows training of models on both data of pure components and of fuels and does not require mixing rules for the calculation of the bulk property. This drastically increases the number of available training data and allows the direct learning of the mixing behavior. For the modeling, we use a Monte-Carlo dropout neural network, a probabilistic machine learning algorithm, that estimates prediction uncertainties due to possible unidentified isomers and dissimilarity of training and test data. Models are developed to predict the freezing point, flash point, net heat of combustion, and temperature-dependent properties such as density, viscosity, and surface tension. We investigate the effect of the presence of fuels in the training data on the predictions for up to 82 conventional fuels and 50 synthetic fuels. The results of the predictions are compared on three metrics that quantify accuracy, precision, and reliability. These metrics allow a comprehensive estimation of the predictive capability of the models. For the prediction of density, surface tension, and net heat of combustion, the M-QSPR method yields highly accurate results even without the presence of fuels in the training data. For properties with nonlinear behavior over temperature and complex fuel component interactions, like viscosity and freezing point, the presence of fuels in the training data was found to be essential for the method.

QSPR Models for the Molar Refraction, Polarizability and Refractive Index of Aliphatic Carboxylic Acids Using the ZEP Topological Index

The molar refraction, polarizability, and refractive index for a series of monocarboxylic, dicarboxylic, and unsaturated monocarboxylic acids, having a symmetric or asymmetric structure, were investigated by the application of quantitative structure property relationship (QSPR) technique. We used a linear regression method and a single molecular descriptor, the ZEP topological index, calculated in a simple manner, with the help of weighted electronic distances, and also calculated on the basis of the chemical structure of the molecules. The high-quality performance and predictive ability of the QSPR models obtained were validated by means of specific validation techniques: y-randomization test, the leave-one-out cross validation procedure, and external validation. The investigated properties are well modeled (with r2 > 0.99) by the ZEP index, using the regression analysis as a statistical tool for developing reliable QSPR models. Our approach provides an alternative technique to the existing additive methods for predicting the molar refraction and polarizability of carboxylic acids, which is essentially based on the summation of atom and/or functional group contributions or bond contributions, and of some correction increments.

Prediction of degradability of micropollutants by sonolysis in water with QSPR - a case study on phenol derivates

The increasing quantity and variety of organic contaminants discharged into surface and groundwater increase the necessity of additional and suitable water treatment methods, which can be incorporated into existing wastewater treatment plants. The huge variety of micropollutants and local variability of the composition of the organic load or matrix effects paired with multiple possible degradation processes lead to the requirement of a recommendation tool for the best possible water treatment method under given local conditions. Due to the diversity of physicochemical properties of micropollutants, such predictions are challenging. In this study, a quantitative correlation between the structural properties of certain micropollutants and their degradability using high-frequency sonolysis has been investigated. Therefore, Quantitative Structure-Property Relationship (QSPR) has been applied on a set of phenol derivates. To obtain the kinetic data, all experiments have been conducted in standardized, constant conditions for all 32 investigated phenol derivates. QSPR modelling was then executed using the software PaDEL for descriptor calculation and the software QSARINS for the overall modelling process including genetic algorithm (GA) and multiple linear regression (MLR). The final model consisting of 5 molecular descriptors was selected using a multi-criteria decision-making method based on extensive statistical parameters. The predictive power and robustness of the model was evaluated by means of internal cross validation and external validation using an independent validation set. The final selected model showed very good values for regression abilities, predictive power as well as stability (R2adj = 0.9455, CCCtr = 0.9777, Q2loo = 0.9285, CCCext = 0.9797 and Q2ext-F1 = 0.9711). The applicability domain of the QSPR model was defined based on the Williams plot and Insubria plot. The five OECD principles for the application of QSPR/QSAR modelling in industry and regulation were fulfilled in the whole process to the best of our knowledge, including the collection of the underlying experimental data as well as the entire modelling process. The final QSPR model included the molecular polarity and occurrence of hydrogen bonds as major influences on the reaction rate constants in accordance with previous studies. Nevertheless, potential biases in the selection of these descriptors due to the small size of the dataset were highlighted.

Can deep learning algorithms enhance the prediction of solute descriptors for linear solvation energy relationship approaches?

Experimental solute descriptors for about 8,000 chemicals are currently available to apply physicochemical property predictions based on linear solvation energy relationship (LSER) models. The solute descriptors can be predicted by fragmental-based quantitative structure-property relationship (QSPR) models. However, the predictions are problematic for larger chemical structures, including multiple functional groups. We developed deep neural networks (DNNs) as alternative prediction models based on graph representations of the chemicals. The root mean square errors rmses range between 0.11 and 0.46 for the different solute descriptors. The predictions of the solute descriptors were compared to predictions from the QSPR of LSERD (an online database) and ACD/Absolv (a commercial software). We further investigated the predictive power of all tools based on three different datasets of experimentally determined partition coefficients, namely the octanol-water partition coefficient (Kow), the octanol-air partition coefficient (Koa), and the water-air partition coefficient (Kwa). Additionally, we used two different sets of retention data for GC and LC to evaluate the results of all prediction tools. All prediction tools perform comparably well with rmses of ∼ 1.0 log unit for the Kow dataset (12,010 chemicals) and ∼ 1.3 log units for the Kwa dataset (696 chemicals), for example. Nevertheless, larger chemical structures are predicted poorly by each approach. We recommend to use the novel DNN model as a complementary prediction tool.

Anticorrosive Properties of Theophylline on Aluminium Corrosion in 1 M HCl: Experimental, and Computational Assessment of Iodide Ions Synergistic Effect

Aims : The aim of this study is to show the anti-corrosive properties of 1,3- dimethyl-7H-purine-2,6-dione on aluminium corrosion in 1 M hydrochloric acid and to study the synergy effect between iodide ions and this molecule. Background: There is a need for a research on eco-friendly, low toxic, and biodegradable corrosion inhibitors capable of protecting metals in order to support industrialists who spend large sums of money on replacing their corroded equipment. Objective: The main objective is to study the anti-corrosive properties of theophylline on aluminium corrosion in 1 M HCl. Methods: The anti-corrosive properties of theophylline on aluminium corrosion in 1 M HCl were evaluated using mass loss, Density Functional Theory at B3LYP/6-31G (d), and Quantitative Structure-Property Relationship methods. Results: The results obtained show that theophylline inhibition efficiency increases with concentration but decreases with increasing temperature with a maximum value of 88% for 5.10-3M at T = 298 K. The result from absorption isotherms reveals that theophylline adsorbs to the aluminium surface according to the modified Langmuir isotherm. Adejo Ekwenshi's isotherm has shown that the molecule adsorption on aluminium is essentially of a physical nature. Thermodynamic adsorption and activation parameters were calculated and analyzed. A synergistic effect between the studied molecule and the iodide ions was found. Furthermore, global and local reactivity were analyzed through density functional theory calculations. The quantitative structure-property relationship model has been used to correlate experimental and theoretical inhibition efficiencies. Conclusion: Theophylline is an excellent aluminum corrosion inhibitor in the current studied solution. Theoretical results were in agreement with the experimental data. Other: Theophylline is an excellent aluminum corrosion inhibitor in the current studied solution. Theoretical results were in agreement with the experimental data. Other: Finally, the best set of parameters to model the inhibition efficiency of analogous molecules were found.

Comparison of Machine Learning Methods towards Developing Interpretable Polyamide Property Prediction.

Polyamides are often used for their superior thermal, mechanical, and chemical properties. They form a diverse set of materials that have a large variation in properties between linear to aromatic compounds, which renders the traditional quantitative structure–property relationship (QSPR) challenging. We use extended connectivity fingerprints (ECFP) and traditional QSPR fingerprints to develop machine learning models to perform high fidelity prediction of glass transition temperature (), melting temperature (), density (), and tensile modulus (). The non-linear model using random forest is in general found to be more accurate than linear regression; however, using feature selection or regularization, the accuracy of linear models is shown to be improved significantly to become comparable to the more complex nonlinear algorithm. We find that none of the models or fingerprints were able to accurately predict the tensile modulus , which we hypothesize is due to heterogeneity in data and data sources, as well as inherent challenges in measuring it. Finally, QSPR models revealed that the fraction of rotatable bonds, and the rotational degree of freedom affects polyamide properties most profoundly and can be used for back of the envelope calculations for a quick estimate of the polymer attributes (glass transition temperature, melting temperature, and density). These QSPR models, although having slightly lower prediction accuracy, show the most promise for the polymer chemist seeking to develop an intuition of ways to modify the chemistry to enhance specific attributes.

Computation of Vertex-Based Topological Descriptors of Organometallic Monolayers of TM 3 C 12 S 12

Topological descriptors are mathematical values related to chemical structures which are associated with different physicochemical properties. The use of topological descriptors has a great contribution in the field of quantitative structure-property relationship (QSPR) and quantitative structure-activity relationship (QSAR) modeling. These are mathematical relationships between different molecular properties or biological activity and some other physicochemical or structural properties. In this article, we calculate few vertex degree-based topological indices/descriptors of the organometallic monolayer structure. At present, the numerical programming of the biological structure with topological descriptors is increasing in consequence in invigorating science, bioinformatics, and pharmaceutics.