Group Contribution Method Research Articles

An Artificial Intelligence Approach to Predict Physical Properties of Liquid Hydrocarbons

Accurate physical property prediction of newly developed compounds is vital across various industrial sectors, particularly for the customization of fuels and additives. Artificial intelligence (AI) has recently emerged as a best practice in numerous industrial fields because of its capacity for swift and precise calculations. While conventional methods such as group contribution models have been used to estimate physical properties from molecular structure, AI offers significant potential for improving the predictive accuracy. Thus, this work focuses on developing an AI model to predict key properties – boiling points, melting points, and flashpoints – of various hydrocarbons, to demonstrate the AI's superior predictive capabilities. A dataset consisting of 202 organic compounds was created and multilayer perceptron (MLP) neural networks were employed to estimate these properties using atomic numbers, functional groups, and molecular complexity as inputs. The model's performance was evaluated and compared against conventional group contribution methods on the same dataset. The AI model was further tested on new acetal compounds, revealing its broader applicability in both fuel and chemical sectors. Results show that the AI outperformed conventional methods, excelling in 5 out of 8 hydrocarbon types for boiling points, 7 for melting points, and all 8 for flashpoints.

Group Contribution Revisited: The Enthalpy of Formation of Organic Compounds with “Chemical Accuracy” Part VI

In this paper we provide the reader with a ready to use Group Contribution (GC) method for the heat of formation (gaseous state) of organics in the form of an Excel spreadsheet with all data, enabling further predictions, and an accompanying manual on how to use the GC model for predicting the heat of formation for organics. In addition, in order to widen the applicability of the method whilst retaining chemical accuracy compared to our previous publications on this topic, we include further chemical groups including acetals, benzyl ethers, bicyclic hydrocarbons, alkanediols and glycerol, polycyclic aromatic hydrocarbons, aromatic fluoro compounds, and finally several species which we include to illustrate how the GC model can be successfully applied to species we did not consider during the parameterization of the GC model parameters.

Removal of Per- and Polyfluoroalkyl Substances (PFAS) in fixed bed anion exchange reactors: Factors determining PFAS uptake capacity and models predicting PFAS breakthrough

Per- and polyfluoroalkyl substances (PFAS) are widespread contaminants with adverse environmental and public health effects. Anion exchange (IX) processes can effectively remove many PFAS from water. Objectives of this research were to (1) quantify the effects of PFAS structure and background water matrix constituents [dissolved organic matter (DOM) and major inorganic anions (bicarbonate, chloride, sulfate, and nitrate)] on PFAS uptake capacity of IX resins (KIX), and (2) develop models that predict PFAS breakthrough in packed bed IX columns from PFAS structure and background water matrix characteristics. Rapid small-scale column tests were conducted with two IX resins and 18 water matrices to determine breakthrough curves and KIX values for 23 PFAS. A quantitative structure-property relationship (QSPR) was developed that predicts KIX for different PFAS from a newly developed concept, the normalized chain length (NCL). NCL for each PFAS was calculated using a group contribution method, in which the contributions of PFAS structural fragments to NCL were normalized to that of -CF2-, which was assigned a value of 1.00. Relative to -CF2-, contributions of the structural fragments -O-, -CH2-, >CF-CF3, -CHF-, and -SO3H to NCL for the purposes of KIX prediction were 0.34, -1.87, 1.70, 0.00, and 3.65, respectively. DOM and nitrate adversely impacted PFAS removal, while chloride, sulfate, and bicarbonate, when added at concentrations of up to 3 meq/L above native background levels, had negligible effects on KIX. Based on the obtained water matrix effects, pseudo-single solute models were developed for the two tested IX resins, with which PFAS breakthrough curves for fixed bed IX columns can be predicted using only the NCL and the total organic carbon and nitrate concentrations of the influent water.

Modelling of wall-bounded cavitating flow and spray mixing in multi-component environments using the PC-SAFT equation of state

This work introduces a numerical multiphase model for multi-component mixtures, utilizing tabulated data for physical and transport properties across a spectrum of conditions from near-vacuum pressures to supercritical states. The property data are derived using Perturbed Chain Statistical Associating Fluid Theory (PC-SAFT), vapor-liquid equilibrium (VLE) calculations, entropy scaling methodologies, and Group Contribution (GC) methods. These techniques accurately reflect the thermodynamic behaviors of real fluids, avoiding the empirical estimation of Equation of State (EoS) input parameters. Implemented in OpenFOAM, the fluid dynamics solver is designed to address the three-dimensional Navier-Stokes equations for multi-component mixtures. The methodology integrates operator splitting to manage hyperbolic and parabolic steps distinctively. Hyperbolic terms are solved using the HLLC (Harten-Lax-van Leer-Contact) solver with temporal integration performed via a third-order Strong-Stability-Preserving Runge–Kutta (SSP-RK3) method. Viscous stress tensor contributions in the momentum equation are handled through an implicit velocity correction equation, while parabolic terms in the energy equation are explicitly solved. The simulation efficiency is further enhanced by adaptive Local Time Stepping and the Immersed Boundary (IB) method, which addresses interactions between the fluid and solid boundaries. Turbulence is resolved using the Wall Adaptive Large Eddy (WALE) model. Applied to high-pressure diesel fuel spray injections into non-reacting (nitrogen) gas environments, the model has been validated against Engine Combustion Network (ECN) data for the Spray-C configuration, featuring a fully cavitating multi-hole orifice. Results demonstrate that the model achieves accurate predictions across a broad range of tested conditions without the need for tuning or calibration parameters.

Simulation of Gas‐Particle Partitioning of Semi‐Volatile n‐Alkanes and PAHs in Nanjing, China, and Denver, United States: Effects of Vapor Pressure and Surface Adsorption Estimation

AbstractThe concentration data of semi‐volatile n‐alkanes and PAHs in the gas phase and in PM2.5 were obtained from Nanjing and Denver. First, the gas‐particle partitioning coefficients of the target compounds were calculated and compared with the predictions based on the equilibrium absorptive partitioning theory. Although the vapor pressure (poL) estimation method was selected to improve the agreement between measured and predicted partitioning coefficients, n‐alkanes in Denver and PAHs in both cities exhibited stronger sorption to PM2.5 than predicted. By including the adsorption mechanism, the average logarithms and temporal variations of the partitioning coefficients were well simulated in both Nanjing and Denver. The partitioning of n‐alkanes in Nanjing can be primarily explained by the absorption mechanism, while the adsorption mechanism dominates that for n‐alkanes in Denver and PAHs in both cities. To obtain an optimal simulation, the poL of n‐alkanes and PAHs was estimated using the SPARC and SIMPOL—two group contribution methods, respectively, and the difference between desorption and vaporization enthalpies was set between 2 and 3 kcal mol−1 for n‐alkanes in Denver and PAHs in both cities. Therefore, the estimation of poL and surface adsorption should be carefully considered when parameterizing the gas‐particle partitioning of non‐polar organic compounds in future field and modeling studies.

Prediction of Thermochemical Properties of Long-Chain Alkanes Using Linear Regression: Application to Hydroisomerization.

Linear regression (LR) is used to predict thermochemical properties of alkanes at temperatures (0-1000) K to study chemical reaction equilibria inside zeolites. The thermochemical properties of C1 until C10 isomers reported by Scott are used as training data sets in the LR model which is used to predict these properties for alkanes longer than C10 isomers. Second-order groups are used as independent variables which account for the interactions between the neighboring groups of atoms. This model accurately predicts Gibbs free energies, enthalpies, Gibbs free energies of formation, and enthalpies of formation for alkanes which exceeds the chemical accuracy of 1 kcal/mol and outperforms the group contribution methods developed by Benson et al., Joback and Reid, and Constantinou and Gani. Predictions from our model are used to compute the reaction equilibrium distribution of hydroisomerization of C10 and C14 isomers in MTW-type zeolite. Calculation of reaction equilibrium distribution inside zeolites also requires Henry coefficients of the isomers which can be computed using classical force field-based molecular simulations using the RASPA2 software for which we created an automated workflow. The reaction equilibrium distribution for C10 isomers obtained using the LR model and the training data set for this model are in very good agreement. The tools developed in this study will enable the computational study of hydroisomerization of long-chain alkanes (>C10).

Ionic liquids based azeotropic separation of refrigerants R410A (R32+R125) and R508B (R23+R116) through hybrid extractive distillation

ABSTRACT The increment in GreenHouse Gases (GHGs) promotes high Global Warming Potential (GWP) due to the production and consumption of hydrofluorocarbons (HFCs) that diverts the aim toward the recycling and reclaiming of refrigerant gases. However, the presence of azeotrope in refrigerant mixtures restricts the recycling process. Therefore, in this work, the refrigerant R410A and R508B have been separated through the hybrid extractive distillation method by using [HMIM][TF2N] and [BMIM][BF4] as a solvent in ASPEN Plus. UNIFAC model has been used for the correlation of ionic liquid in ASPEN Plus. For thermodynamic properties, Cosmotherm has been used with the collaboration of TurbromoleX. The Vapor-liquid Equilibrium (VLE) data has been predicted by using the UNIFAC group contribution method. The Total Annual Cost (TAC) analysis with optimization has been performed. Besides that, the energy analysis has also been performed. In the overall separation process and (TAC) analysis, [BMIM][BF4] proves to be more efficient with high separation efficiency and minimal TAC requirements. Whereas, [HMIM][TF2N]/R508B provides less CO2 emission.

Modeling the thermodynamic properties of cyclic alcohols with the SAFT-γ Mie approach: application to cyclohexanol and menthol systems

Cyclic alcohols are commonly found in natural and synthetic products and are involved in many biological processes. An ability to model their thermodynamic properties is of interest in food, flavouring, and pharmaceutical manufacturing, particularly for mixtures for which available experimental data are limited. Good examples are mixtures including menthol, a naturally occurring cyclic alcohol widely used in the food and pharmaceutical industries, well-known for its cooling-sensation properties and its role in the discovery of the TRPM8 receptor. Here, we extend the SAFT-γ Mie group-contribution method to model cyclic alcohols, by introducing a new cCHOH group (c for cyclic) composed of two identical Mie segments. Three association sites (two electronic sites of type e and one hydrogen site of type H) are also included to mediate hydrogen bonding. New parameters that characterise the group interactions (one new like interaction and 26 new unlike interactions) are developed and are employed to determine the thermophysical properties (vapour pressure, saturation density, vaporisation enthalpy, and second-order thermodynamic derivative properties) of pure cyclic alcohols, and mixture properties (vapour–liquid equilibria, liquid–liquid equilibria, solid–liquid equilibria, density, and excess enthalpy) of cyclic alcohols in several solvents. The quality of the model is evaluated by comparing predictive calculations with experimental data of 76 systems: six pure fluids and 70 binary mixtures, selected from a large range of solvent families: cyclic, linear, and branched alkanes, 1-alcohols, 2-alcohols, 2-ketones, esters, aromatic compounds, water, and carbon dioxide. Very good overall agreement is found, including for the prediction of solid–liquid solubility, which confirms the transferability of the new group parameters. Together with previously developed parameters, these open the way for the prediction of the thermodynamic properties of further complex mixtures.

Thermophysical Characterization of Protic Ionic Liquids: Density Measurements, Group Contribution Method, and PC-SAFT Modeling Application

Thermophysical Characterization of Protic Ionic Liquids: Density Measurements, Group Contribution Method, and PC-SAFT Modeling Application

Modeling Dipolar Molecules with PCP-SAFT: A Vector Group-Contribution Method.

Predicting thermodynamic equilibrium properties is essential to develop chemical and energy conversion processes in the absence of experimental data. For the modeling of thermodynamic properties, statistical associating fluid theory (SAFT)-based equations of state, such as perturbed-chain polar (PCP)-SAFT, have been proven powerful and found broad application. The PCP-SAFT parameters can be predicted by group-contribution (GC) methods. However, their application to the dipole term is substantially limited: current GC methods neglect the dipole term or only allow for a single dipolar group per substance to avoid handling the molecular dipole moment's symmetry effects. Still, substances with multiple dipolar groups are highly relevant, and their description substantially improves by including the dipole term in SAFT models. To overcome these limitations, this work proposes a vector-addition-based (Vector-)GC method for the dipole term of PCP-SAFT that accounts for molecular symmetry. The Vector-GC employs information on the substance's molecular 3D structure to predict the molecular dipole moment through a vector addition of bond contributions. Combining the proposed sum rule for dipole moments with established sum rules for the remaining parameters yields a consistent GC method for PCP-SAFT for dipolar substances. The prediction capabilities of the Vector-GC method are analyzed against experimental data for two substance classes: nonassociating oxygenated and halogenated substances. We demonstrate that the Vector-GC method improves vapor pressure and liquid density predictions compared to neglecting the dipole term. Moreover, we show that the Vector-GC method enables differentiation between cis- and trans-isomers. The Vector-GC method, hence, substantially increases the predictive capabilities and applicability domain of GC methods. All parameters are provided as JSON and CSV files, and the Vector-GC method is available through an open-source python package. Additionally, the developed regression framework for GC methods for PCP-SAFT is openly available. The regression framework can be employed to regress the Vector-GC method to other substance classes and is easily adaptable to other sum rules for PCP-SAFT.

High-temperature heat pump working fluid development based on molecular design and the group contribution method

High-temperature heat pumps are widely recognized as a critical approach for decarbonizing energy-intensive industries, and their performance is significantly influenced by the working fluids. However, conventional working fluid selection methods primarily focus on existing fluids, limiting the selection scale. Therefore, a group-contribution-based computer-aided molecular design method has been proposed to design and select working fluids for high-temperature heat pumps on a large scale, comprehensively considering environmental impact, safety, operating reliability and performance. A total of 9771 working fluids were generated by the computer-aided process based on the backtracking searching algorithm. Under different working conditions, 208, 232, 243, 228, and 208 working fluids have passed the multi-level screening process based on the group contribution method, respectively, with an overall passing rate below 2.49 %. The H(C)FOs among the top 50 working fluids, ranked by the coefficient of performance, were further analyzed. Statistically, 92.4 % and 98.73 % of the total top 50 working fluids had their Tbr and Tc values within the ranges of 0.640–0.675 and 432–480 K, respectively. Among the 18 HCFOs within the top 50 list, 83.3 % were non-existent. R1251zd and R1344zf demonstrated the best ranking results, while R1242zf, R1233yf, and R1233xf exhibited superior working condition flexibility. As an existing HCFO, R1251yd outperformed other working fluids. Of the 35 HFOs in the top 50 list, 88.57 % were non-existent, and R1381myz, R1371my and R1381mzz showed both exceptional performance and superior working condition adaptability. In conclusion, there remain promising prospects and a notable necessity for the new H(C)FO development, and guidance can be expected from this study in this regard.

Using the group contribution method to predict the flash temperature of biodiesel and ethanol mixtures

This article presents a novel approach to predict the flash temperature of biodiesel and ethanol mixtures using the Group Contribution Method (GCM). Expanding on the pioneering work by Liaw et al. (2003), our method employs GCM to calculate the activity coefficients of biodiesel and ethanol components in the mixture. Estimating these coefficients, crucial for accurate flash temperature prediction, involves a comprehensive analysis of composition, functional groups, and vapor-liquid equilibrium (VLE) data. For this purpose, the composition of the mixture components in biodiesel, the functional groups within each biodiesel component, the composition ratios of biodiesel and ethanol in the mixture, and the functional groups present in ethanol are considered. Given that the use of UNIQUAC and NRTL models requires estimating adjustable parameters, VLE data for ethanol and biodiesel mixtures are employed to calculate the activity coefficients. This approach not only aids in estimating these coefficients but also facilitates determining the values associated with each functional group. Flash temperature predictions for biodiesel and ethanol mixtures obtained through various models, including the ideal solution, UNIQUAC, NRTL, and our proposed GCM, are rigorously assessed. The results indicate that the GCM method outperforms the alternatives, exhibiting the lowest error with a deviation of just 1.72 K compared to deviations of 1.77 K, 1.75 K, and 1.73 K for the ideal solution, UNIQUAC, and NRTL models, respectively. This research offers a promising approach for flash point estimation in complex systems, such as biodiesel-ethanol blends, contributing to the ongoing exploration in this field.

Prediction of ionic liquid surface tension via a generalized interpretable Structure‐Surface Tension Relationship model

AbstractIonic liquids' (ILs) surface tension, vital in liquid interface research, faces challenges in measurement methods—time‐consuming and labor‐intensive. The Structure‐Surface Tension Relationship (SSTR) is crucial for understanding the surface tension laws of ionic liquids, helping to predict surface tension and design ionic liquids that meet target requirements. In this study, SMILES string and group contribution methods were used to generate descriptors, and the random forest and multi‐layer perceptron (MLP) models were cross combined with the two descriptor generation methods to establish the SSTR model, providing a comprehensive framework for predicting the surface tension of ionic liquids. String‐MLP excels with high accuracy (R2 = 0.995, RMSE = 0.686, AARD% = 0.71%) for diverse ILs' surface tension values. Meanwhile, the Shapley Additive exPlanning (SHAP) method was used to test the impact of different features on model prediction, increasing the transparency and interpretability of the model.

Modelling the high rate and shock response of polymers from a molecular basis

A brief review is presented of current capability in the prediction of the thermomechanical properties of polymers with a focus on high rate and shock response. Group contribution methods have demonstrated that accurate predictions of macroscopic properties can be made from knowledge of molecular architecture alone without the need to fit to experiment. Properties are dominated by polymers’ changing numbers of molecular vibrations and the mechanical loss associated with this process. Areas for beneficial research would be: to improve understanding of the glass transition at pressures above 1 GPa; construction of a complete equation of state based on group contribution methods in order to avoid the need for fitting; accurate implementation schemes for relaxation-based hyperelasticity.

Research on flammability of 3,3,3-trifluoropropene and its binary mixtures with flame retardants

3,3,3-trifluoropropene (R1243zf), which has good environmental and cycling characteristics, is a very promising fourth generation new working fluid. However, it is flammable, so studying its combustion characteristics and how to suppress its flammability are of great significance. Firstly, the flammability limits of three binary mixtures composed of a combustible substance (R1243zf) and three flame retardants (R245fa, R1216, and R13I1) were experimentally studied. The results showed that as the proportion of flame retardants increased, the lower flammability limit of the mixture increased, while the upper flammability limit decreased except for R245fa. The critical suppression ratios of R245fa, R1216, and R13I1 on R1243zf were 4, 0.538, and 0.538, respectively. Then, the improved group contribution method was used to predict the minimum inhibitory concentration of R1243zf, with an error of only 4.57 % compared to the experimental values. Finally, the influence of variable ambient temperature (25–100℃) on the flammability of R1243zf was studied, and the calculated adiabatic flame temperature method was used to predict the flammability limit at different temperatures. The error between the predicted values and the experimental values is only 1.7 %. The relevant results are of great significance for the safe application of R1243zf and new binary mixed working fluids.

Modeling Study on Heat Capacity, Viscosity, and Density of Ionic Liquid–Organic Solvent–Organic Solvent Ternary Mixtures via Machine Learning

Physicochemical properties of ionic liquids (ILs) are essential in solvent screening and process design. However, due to their vast diversity, acquiring IL properties through experimentation alone is both time-consuming and costly. For this reason, the creation of prediction models that can accurately forecast the characteristics of IL and its mixtures is crucial to their application. This study proposes a model for predicting the three important parameters of the IL-organic solvent–organic solvent ternary system: density, viscosity, and heat capacity. The model incorporates group contribution (GC) and machine learning (ML) methods. A link between variables such as temperature, pressure, and molecular structure is established by the model. We gathered 2775 viscosity, 6515 density, and 1057 heat capacity data points to compare the prediction accuracy of three machine learning methods, namely, artificial neural networks (ANNs), extreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM). As can be observed from the findings, the ANN model produced the best results out of the three GC-based ML methods, even though all three produced dependable predictions. For heat capacity, the mean absolute error (MAE) of the ANN model is 1.7320 and the squared correlation coefficient (R2) is 0.9929. Regarding viscosity, the MAE of the ANN model is 0.0225 and the R2 is 0.9973. For density, the MAE of the ANN model is 7.3760 and the R2 is 0.9943. The Shapley additive explanatory (SHAP) approach was applied to the study to comprehend the significance of each feature in the prediction findings. The analysis results indicated that the R-CH3 group of the ILs, followed by the imidazolium (Im) group, had the highest impact on the heat capacity property of the ternary system. On the other hand, the Im group and the R-H group of ILs had the most effects on viscosity. In terms of density, the Im group of the ILs had the greatest effect on the ternary system, followed by the molar fraction of the organic solvent.

Screening HFC/HFO and ionic liquid for absorption refrigeration at the atomic scale by the prediction model of machine learning

Absorption refrigeration is a highly effective method for utilizing renewable energy, as it can be driven by low-grade heat sources such as industrial waste heat, solar energy, and geothermal energy. The development of new working pairs, particularly hydrofluorocarbon/hydrofluoroolefin refrigerants combined with ionic liquids, has been pivotal in enhancing the cooling efficiency of absorption refrigeration systems. These systems rely on the solubility difference between the generator and absorber, making solubility a crucial factor in determining their efficiency. In this context, we have established an advanced solubility estimation model. This model employs the Attention E(n)-equivariant Graph Neural Network (AEGNN) applied to disconnected graphs, enabling comprehensive learning from both topological and Euclidean structural information. Our atomic-scale model demonstrates significantly higher accuracy than traditional group contribution methods, with an average absolute deviation of 0.003 mol/mol from experimental data. Moreover, it encompasses a much broader range of working pairs. Through extensive screening, we have identified working pairs with high estimated solubility differences. Compared to the high-efficiency working pair identified in the literature, the best-screened working pairs exhibit an improvement in solubility differences by more than 0.3 mol/mol under common operating conditions.

Synthesis, characterization and film preparation of 6FDA-based polyimides with polar groups arranged for thermal crosslinking

Abstract Thermal crosslinking of polyimides is a relevant methodology to increase their chemical resistance. This reaction is induced by heating after polymerization. Although this reaction is considered a viable route to modify polyimides, there is controversy over the specific temperature for the reaction. To contribute to the knowledge of crosslinked structures, we performed a differential scanning calorimetric (DSC) and solubility test studies on a series of polyimides with different molar ratios of carboxylic acid groups. With 3,5-diaminobenzoic acid (DABA) as a source of carboxylic acid groups, the copolyimides were synthesized from 2,2′-bis(3,4-dicarboxyphenyl)hexafluoropropane dianhydride (6FDA) and 4,4ʹ-hexafluoropropylidenebis(p-phenylenoxy)dianiline (6FPA) to form blocks of 6FDA:6FPA (n) and 6FDA:DABA (m) from m = 0–100 %. Flat films were prepared at 50 ± 3 µm with polyimides. In the thermogram of the flat films, a heat absorption zone associated with the decarboxylation reaction was identified with an endothermic minimum at 260 °C, which was assigned to the crosslinking temperature. Furthermore, heating time was varied for 3–36 h to determine the insolubility of the thermally crosslinked polyimide; from 24 h of heating the polyimide were insoluble. The polyimides were thoroughly characterized by ATR-FTIR, 1H NMR, TGA, SEM, the fractional free volume (FFV) was determined by the Bondi’s group contribution method and solubility to different organic solvents. The FFV of crosslinked polyimides increases relative to pristine polyimides by 2 % for those with higher DABA content and 40 % for those with lower DABA content. Finally, also the glass transition temperature increases in the crosslinked polyimides relative to the pristine ones from 280 °C for the lowest DABA content to 402 °C for the highest DABA content, demonstrating the increase in the thermal resistance of the structure.

Machine learning approach to predict Hansen solubility parameters of cocrystal coformers via integrating group contribution and COSMO-RS

Hansen Solubility Parameters (HSPs) have been a hot topic on predicting the tendency of pharmaceutical cocrystals formation and cocrystal coformers (CCFs) screening. However, the limitation of such application is the lack of models to accurately predict the values of HSPs for drug CCFs with more structural complexity. Accordingly, three ML (machine learning) models, i.e. ANN (Artificial Neural Network), XGBoostRegressor (Extreme Gradient Boosting Regressor) and LGBMRegressor (Light Gradient Boosting Machine Regressor), were developed for predicting the HSPs on CCFs screening for drugs. The HSPs database for 181 CCFs (containing alcohols, alkenes, aromatics, haloalkanes, amines, ketones, ethers, amides, esters, pharmaceuticals, alkanes, acids, nitroalkanes) were established and classified into the training set (140 compounds) and the test set (41 compounds with various functional polarity and groups, covering solid reagents and solvents). The prediction molecular descriptors were combined from the GC (Group Contribution) methods, the COSMO-RS (the Conductor-like Screening Model for Real Solvents) sigma-moments and energy descriptors. The results showed that ANN and XGBoostRegressor beat out LGBMRegressor in predicting HSPs for CCFs. Finally, SHapley Additive exPlanations (SHAP) was employed to visualize and explain the most important characteristics and effects on predicting HSPs via XGBoostRegressor, indicating that CH3, M2 and MHbdon3 had a significant influence and high contribution to the prediction of δd, δp and δh, respectively. The coupled GC and COSMO-RS strategy had been proven as a promising tool to predict HSPs through XGBoostRegressor for screening, designing, and selecting CCFs for drugs.

Multivariate Analysis of Solubility Parameters for Drug-Polymer Miscibility Assessment in Preparing Raloxifene Hydrochloride Amorphous Solid Dispersions.

The success of obtaining solid dispersions for solubility improvement invariably depends on the miscibility of the drug and polymeric carriers. This study aimed to categorize and select polymeric carriers via the classical group contribution method using the multivariate analysis of the calculated solubility parameter of RX-HCl. The total, partial, and derivate parameters for RX-HCl were calculated. The data were compared with the results of excipients (N = 36), and a hierarchical clustering analysis was further performed. Solid dispersions of selected polymers in different drug loads were produced using solvent casting and characterized via X-ray diffraction, infrared spectroscopy and scanning electron microscopy. RX-HCl presented a Hansen solubility parameter (HSP) of 23.52 MPa1/2. The exploratory analysis of HSP and relative energy difference (RED) elicited a classification for miscible (n = 11), partially miscible (n = 15), and immiscible (n = 10) combinations. The experimental validation followed by a principal component regression exhibited a significant correlation between the crystallinity reduction and calculated parameters, whereas the spectroscopic evaluation highlighted the hydrogen-bonding contribution towards amorphization. The systematic approach presented a high discrimination ability, contributing to optimal excipient selection for the obtention of solid solutions of RX-HCl.