A variational heteroencoder based on recurrent neural networks, trained with SMILES linear notations of molecular structures, was used to derive the following atomic descriptors: delta latent space vectors (DLSVs) obtained from the original SMILES of the whole molecule and the SMILES of the same molecule with the target atom replaced. Different replacements were explored, namely, changing the atomic element, replacement with a character of the model vocabulary not used in the training set, or the removal of the target atom from the SMILES. Unsupervised mapping of the DLSV descriptors with t-distributed stochastic neighbor embedding (t-SNE) revealed a remarkable clustering according to the atomic element, hybridization, atomic type, and aromaticity. Atomic DLSV descriptors were used to train machine learning (ML) models to predict 19F NMR chemical shifts. An R2 of up to 0.89 and mean absolute errors of up to 5.5 ppm were obtained for an independent test set of 1046 molecules with random forests or a gradient-boosting regressor. Intermediate representations from a Transformer model yielded comparable results. Furthermore, DLSVs were applied as molecular operators in the latent space: the DLSV of a halogenation (H→F substitution) was summed to the LSVs of 4135 new molecules with no fluorine atom and decoded into SMILES, yielding 99% of valid SMILES, with 75% of the SMILES incorporating fluorine and 56% of the structures incorporating fluorine with no other structural change.

QSPR mathematically links physicochemical properties with the structure of a molecule. The physicochemical properties of chemical molecules can be predicted using topological indices. It is an effective method for eliminating costly and time-consuming laboratory tests. We established a QSPR between mev-degree and mve-degree-based indices and the physical properties of benzenoid hydrocarbons. To compute these indices, we designed a program using Maple software and the correlation between indices and physical properties was developed using the SPSS software. Our study reveals that the mve-degree-based sum-connectivity ( χ mve ) and atom bond connectivity ( A B C mve ) index, mev-degree-based Randić ( R mev ) and Zagreb ( M mev ) index are the three most significant parameters and have good prediction ability for the physicochemical properties. We examined that R mev predicts the molar refractivity and boiling point, χ mve predicts the LogP and enthalpy, A B C mve predicts the molecular weight, M mev predicts the Gibb\'s energy, Pie-electron energy and Henry\'s law. Moreover, we computed the indices for the linear [n]-phenylen.

This study investigates the quantitative structure-property relationship (QSPR) modeling of guar gum biomolecules, focusing on their structural parameters. Guar gum, a polysaccharide with diverse industrial applications, exhibits various properties such as viscosity, solubility, and emulsifying ability, which are influenced by its molecular structure. In this research, M -polynomial and associated topological indices are employed as structural descriptors to represent the molecular structure of guar gum. The M -polynomial and associated topological indices capture important structural features, including size, shape, branching, and connectivity. By correlating these descriptors with experimental data on guar gum properties, predictive models are developed using regression analysis techniques. The analysis revealed a strong correlation between the boiling point and molecular weight and all the considered topological descriptors. The resulting models offer insights into the relationship between guar gum structure and its properties, facilitating the optimization of guar gum production and application in various industries. This study demonstrates the utility of M -polynomial and QSPR modeling in elucidating structure-property relationships of complex biomolecules like guar gum, contributing to the advancement of biomaterial science and industrial applications.

This study describes the development and evaluation of six new models for predicting physical-chemical (PC) properties that are highly relevant for chemical hazard, exposure, and risk estimation: solubility (in water SW and octanol SO), vapor pressure (VP), and the octanol-water (KOW), octanol-air (KOA), and air-water (KAW) partition ratios. The models are implemented in the Iterative Fragment Selection Quantitative Structure-Activity Relationship (IFSQSAR) python package, Version 1.1.0. These models are implemented as Poly-Parameter Linear Free Energy Relationship (PPLFER) equations which combine experimentally calibrated system parameters and solute descriptors predicted with QSPRs. Two other ancillary models have been developed and implemented, a QSPR for Molar Volume (MV) and a classifier for the physical state of chemicals at room temperature. The IFSQSAR methods for characterizing applicability domain (AD) and calculating uncertainty estimates expressed as 95% prediction intervals (PI) for predicted properties are described and tested on 9,000 measured partition ratios and 4,000 VP and SW values. The measured data are external to IFSQSAR training and validation datasets and are used to assess the predictivity of the models for \"novel chemicals\" in an unbiased manner. The 95% PI intervals calculated from validation datasets for partition ratios needed to be scaled by a factor of 1.25 to capture 95% of the external data. Predictions for VP and SW are more uncertain, primarily due to the challenges in differentiating their physical state (i.e., liquids or solids) at room temperature. The prediction accuracy of the models for log KOW, log KAW and log KOA of novel, data-poor chemicals is estimated to be in the range of 0.7 to 1.4 root mean squared error of prediction (RMSEP), with RMSEP in the range 1.7-1.8 for log VP and log SW. Scientific contributionNew partitioning models integrate empirical PPLFER equations and QSARs, allowing for seamless integration of experimental data and model predictions. This work tests the real predictivity of the models for novel chemicals which are not in the model training or external validation datasets.

Thermoelectric materials with a high Seebeck coefficient, high electrical conductivity, and low thermal conductivity are required to directly and efficiently convert unused heat into electricity. In this study, we construct models predicting the Seebeck coefficient, electrical conductivity, and thermal conductivity using existing material databases. In addition to the ratios of atoms in the crystals and temperature at which the materials are used, the values from the X-ray diffraction (XRD) spectra were used as inputs to represent the crystal structure of the materials. It was confirmed that the constructed models could predict the properties with high accuracy using the X-ray diffraction values. Additionally, using the constructed models, we succeeded in proposing promising new candidate materials with high Seebeck coefficients, high electric conductivities, and low thermal conductivities.

Aim: This study aims to investigate the passive diffusion of protein kinase inhibitors through the blood-brain barrier (BBB) and to develop a model for their permeability prediction. Materials & methods: We used the parallel artificial membrane permeability assay to obtain logPe values of each of 34 compounds and calculated descriptors for these structures to perform quantitative structure-property relationship modeling, creating different regression models. Results: The logPe values have been calculated for all 34 compounds. Support vector machine regression was considered the most reliable, and CATS2D_09_DA, CATS2D_04_AA, B04[N-S] and F07[C-N] descriptors were identified as the most influential to passive BBB permeability. Conclusion: The quantitative structure-property relationship-support vector machine regression model that has been generated can serve as an efficient method for preliminary screening of BBB permeability of new analogs.
[Box: see text].

In the growing landscape of interest in natural surfactants, selecting the appropriate one for specific applications remains challenging. The extensive, yet often unsystematized, knowledge of microbial surfactants, predominantly represented by rhamnolipids (RLs), typically does not translate beyond the conditions presented in scientific publications. This limitation stems from the numerous variables and their interdependencies that characterize microbial surfactant production. We hypothesized that a computational recipe for biosynthesizing RLs with targeted applicational properties could be developed from existing literature and experimental data. We amassed literature data on RL biosynthesis and micellar solubilization and augmented it with our experimental results on the solubilization of triglycerides (TGs), a topic underrepresented in current literature. Utilizing this data, we constructed mathematical models that can predict RL characteristics and solubilization efficiency, represented as logPRL = f(carbon and nitrogen source, parameters of biosynthesis) and logMSR = f(solubilizate, rhamnolipid (e.g. logPRL), parameters of solubilization), respectively. The models, characterized by robust R2 values of respectively 0.581-0.997 and 0.804, enabled the ranking of descriptors based on their significance and impact-positive or negative-on the predicted values. These models have been translated into ready-to-use calculators, tools designed to streamline the selection process for identifying a biosurfactant optimally suited for intended applications.

In this study, we focus on the development of Quantitative Structure-Property Relationship (QSPR) models to predict the critical micelle concentration (CMC) for per- and polyfluoroalkyl substances (PFASs). Experimental CMC values for both fluorinated and non-fluorinated compounds were meticulously compiled from existing literature sources. Our approach involved constructing two distinct types of models based on Support Vector Machine (SVM) algorithms applied to the dataset. Type (I) models were trained exclusively on CMC values for fluorinated compounds, while Type (II) models were developed utilizing the entire dataset, incorporating both fluorinated and non-fluorinated compounds. Comparative analyses were conducted against reference data, as well as between the two model types. Encouragingly, both types of models exhibited robust predictive capabilities and demonstrated high reliability. Subsequently, the model having the broadest applicability domain was selected to complement the existing experimental data, thereby enhancing our understanding of PFAS behaviour.

BACKGROUND: Veterinary antibiotics are chemical compounds used to kill or inhibit the growth of pathogenic bacteria associated with animal diseases. These molecules can be defined by their retention times (tR) in liquid chromatography-mass spectrometry (LC-MS). One strategy to predict the tR of new veterinary antibiotics is the development of predictive quantitative structure-property relationships (QSPRs), which were used in this study.
RESULTS: A database of 122 antibiotics was selected in which the tR was measured using a Hypersil GOLD column. An optimal three-feature model was developed by integrating the unsupervised variable reduction, replacement method variable subset selection, and multiple linear regression. The negligible differences among the coefficient of determination and the root-mean-square error for the training set (R2 = 0.902 and RMSEC = 0.871) and test set (Q2 = 0.854 and RMSEP = 1.064) indicate a stable and predictive model. In a further step, a more in-depth explanation of the mechanism of action of each descriptor in predicting the tR is provided, with the construction of the theoretical chemical space for accurate predictions of new antibiotics.
CONCLUSIONS: The in silico model developed in this work identified three molecular descriptors associated with aqueous solubility, octanol-water partition coefficient, and the presence of negative and lipophilic atom pairs. The QSPR developed here could be implemented by agricultural and food chemists to identify and monitor existing and new antibiotics within the framework of LC-MS. The computational model was developed in accordance with five principles outlined by the Organization for Economic Co-operation and Development. © 2024 Society of Chemical Industry.

Accurately predicting plant cuticle-air partition coefficients (Kca) is essential for assessing the ecological risk of organic pollutants and elucidating their partitioning mechanisms. The current work collected 255 measured Kca values from 25 plant species and 106 compounds (dataset (I)) and averaged them to establish a dataset (dataset (II)) containing Kca values for 106 compounds. Machine-learning algorithms (multiple linear regression (MLR), multi-layer perceptron (MLP), k-nearest neighbors (KNN), and gradient-boosting decision tree (GBDT)) were applied to develop eight QSPR models for predicting Kca. The results showed that the developed models had a high goodness of fit, as well as good robustness and predictive performance. The GBDT-2 model (Radj2 = 0.925, QLOO2 = 0.756, QBOOT2 = 0.864, Rext2 = 0.837, Qext2 = 0.811, and CCC = 0.891) is recommended as the best model for predicting Kca due to its superior performance. Moreover, interpreting the GBDT-1 and GBDT-2 models based on the Shapley additive explanations (SHAP) method elucidated how molecular properties, such as molecular size, polarizability, and molecular complexity, affected the capacity of plant cuticles to adsorb organic pollutants in the air. The satisfactory performance of the developed models suggests that they have the potential for extensive applications in guiding the environmental fate of organic pollutants and promoting the progress of eco-friendly and sustainable chemical engineering.