Development of machine learning model and analysis study of drug solubility in supercritical solvent for green technology development

The current research is focused on development of machine learning model for estimation of pharmaceutical solubility in supercritical CO₂ as the green solvent. The main aim is to assess the suitability of supercritical processing for preparation of nanomedicine. Oxaprozin was taken as model drug for the solubility measurements, and its solubility was determined at different operational conditions by variation of temperature and pressure of the process. Artificial Neural Network (ANN) model was implemented for simulation of the drug solubility, and the best model was obtained with R² greater than 0.99 for the training and validation as well. The tested model was then exploited to understand the process, and it turned out that both pressure and temperature had major and considerable influence on the solubility of Oxaprozin in supercritical carbon dioxide as solvent. However, the effect of pressure was shown to be more significant on the solubility compared to the effect of pressure, which was attributed to the effect of pressure on the density of the supercritical solvent. The developed ANN model was indicated to be robust in estimating the values of drug solubility in wide range of conditions which can save time and cost of the measurements.     

Novel density-based model for the correlation of solid drugs solubility in supercritical carbon dioxide

A novel density-based model derived by a simple modification of the Jouyban et al. model has been proposed to correlate the solubility of solid drugs in supercritical carbon dioxide. The six-parameter model expresses the solubility only as a function of the solvent density and the equilibrium temperature. This model is in contrast to the Jouyban et al. (J. Superiority. Fluids 24 (2002) 19) model, which gives the solubility as a function of the solvent density and the equilibrium temperature and pressure. The performance of the model has been tested on a database of 100 drugs that account for 2891 experimental data points collected from the literature. The comparison in terms of the mean absolute relative deviation for each solid drug and for the entire database between the proposed model and models that have been suggested to be mostly more accurate demonstrates that the proposed model has the best global correlation performance, exhibiting an overall average absolute relative deviation of 8.13%.     

Measurement and correlation of solubility of artemisinin in supercritical carbon dioxide

Artemisinin is an effective antimalarial drug isolated from the herbal medicine Artemisia annua L. Supercritical fluid extraction is an environment-friendly method for the extraction of artemisinin. In this work, the solubility of artemisinin in supercritical carbon dioxide was determined by static method at three temperatures of 313 K, 323 K, 333 K and pressures from 11 to 31 MPa. The range of experimental solubility data was from 0.498 × 10⁻³ to 2.915 × 10⁻³ mol/mol under the above-mentioned conditions. Two density-based models (Chrastil and Mendez–Santiago–Teja models) were selected to correlate the experimental data of this work, and the average absolute relative deviation (AARD) was 8.32% and 8.33%, respectively. The correlation results showed good agreement with the experimental data.     

Measurement and modeling of metoclopramide hydrochloride (anti-emetic drug) solubility in supercritical carbon dioxide

Knowledge of drug solubility data in supercritical carbon dioxide (SC-CO₂) is a fundamental step in producing nano and microparticles through supercritical fluid technology. In this work, for the first time, the solubility of metoclopramide hydrochloride (MCP) in SC-CO₂ was measured in pressure and temperature range of 12 to 27 MPa and 308 to 338 K, respectively. The results represented a range mole fractions of 0.15 × 10^-5 to 5.56 × 10^-5. To expand the application of the obtained data, six semi-empirical models and three models based on the Peng-Robinson equation of state (PR + VDW, PR + WS + Wilson and PR + MHV1 + COSMOSAC) with different mixing rules and various ways to describe intermolecular interactions were investigated. Furthermore, total enthalpy, sublimation enthalpy and solvation enthalpy relevant to MCP solvating in SC-CO₂ were estimated.     

Measurement and correlation of solubility of anthraquinone dyestuffs in supercritical carbon dioxide

Solubility data of 1,4-diaminoanthraquinone (C.I. Disperse Violet 1) and 1,4-bis(ethylamino)anthraquinone (C.I. Solvent Blue 59) in supercritical carbon dioxide (sc-CO₂) have been measured at the temperatures of (323.15, 353.15, and 383.15) K and over the pressure range from (12.5 to 25.0) MPa by a flow-type apparatus. The solubility of two anthraquinone dyestuffs was obtained over the mole fraction ranges of (1.3 to 26.1) · 10⁻⁷ for 1,4-diaminoanthraquinone (C.I. Disperse Violet 1) and (1.1 to 148.5) · 10⁻⁷ for 1,4-bis(ethylamino)anthraquinone (C.I. Solvent Blue 59). The experimental results have been correlated with the empirical equations of Mendez-Santiago–Teja and Kumar–Johnston expressed in terms of the density of sc-CO₂, and also analyzed thermodynamically by the regular solution model with the Flory–Huggins theory and the Peng–Robinson equation of state modified by Stryjek and Vera (PRSV-EOS) with the conventional mixing rules. Good agreement between the experimental and calculated solubilities of the dyestuffs was obtained.     

Solubility enhancement of decitabine as anticancer drug via green chemistry solvent: Novel computational prediction and optimization

Nowadays, supercritical fluid technology (SFT) has been an interesting scientific subject in disparate industrial-based activities such as drug delivery, chromatography, and purification. In this technology, solubility plays an incontrovertible role. Therefore, achieving more knowledge about the development of promising numerical/computational methods of solubility prediction to validate the experimental data may be advantageous for increasing the quality of research and therefore, the efficacy of novel drugs. Decitabine with the chemical formula C₈H₁₂N₄O₄ is a chemotherapeutic agent applied for the treatment of disparate bone-marrow-related malignancies such as acute myeloid leukemia (AML) by preventing DNA methyltransferase and activation of silent genes. This study aims to predict the optimum value of decitabine solubility in CO₂SCF by employing different machine learning-based mathematical models. In this investigation, we used AdaBoost (Adaptive Boosting) to boost three base models such as Linear Regression (LR), Decision Tree (DT), and GRNN. We used a dataset that has 32 sample points to make solubility models. One of the two input features is P (bar) and the other is T (k). ADA-DT (Adaboost Algorithm-Decision Tree), ADA-LR (Adaboost Algorithm-Linear Regresion), and ADA-GRNN (Generative Regression Neural Network) models showed MAE of 6.54 × 10^?5, 4.66 × 10^?5, and 8.35 × 10^?5, respectively. Also, in terms of R-squared score, these models have 0.986, 0.983, and 0.911 scores, respectively. ADA-LR was selected as the primary model according to numerical and visual analysis. Finally, the optimal values are (P = 400 bar, T = 3.38 K × 102, Y = 1.064 × 10^?3 mol fraction) using this model.     

Measurement and correlation of solubility of benzamide in supercritical carbon dioxide with and without cosolvent

The experimental equilibrium solubility of benzamide in supercritical carbon dioxide was measured at temperatures between 308 K and 328 K and for pressures from 11.0 MPa to 21.0 MPa using a dynamic flow method. The effects of three cosolvents - ethanol, acetone and ethylene glycol, were investigated at a cosolvent molar concentration of 3.5%. The results showed that the solubility was enhanced by the presence of the three cosolvents, and ethanol exhibited the highest cosolvent effect. The solubility data in the absence and presence of cosolvents were correlated by two density-based models. The calculated results showed satisfactory agreement with the experimental data.     

Effects of process parameters on supercritical CO2 extraction of total phenols from strawberry (Arbutus unedo L.) fruits: An optimization study

The aim of this work was to optimize total phenolic yield of Arbutus unedo fruits using supercritical fluid extraction. A Box–Behnken statistical design was used to evaluate the effect of various values of pressure (50–300 bar), temperature (30–80°C) and concentration of ethanol as co‐solvent (0–20%) by CO₂ flow rate of 15 g/min for 60 min. The most effective variable was co‐solvent ratio (p<0.005). Evaluative criteria for both dependent variables (total phenols and radical scavenging activity) in the model were assigned maximum. Optimum extraction conditions were elicited as 60 bar, 48°C and 19.7% yielding 25.72 mg gallic acid equivalent (GAE) total phenols/g extract and 99.9% radical scavenging capacity, which were higher than the values obtained by conventional water (24.89 mg/g; 83.8%) and ethanol (15.12 mg/g; 95.8%) extractions demonstrating challenges as a green separation process with improved product properties for industrial applications.     

Estimating the domain of applicability for machine learning QSAR models: a study on aqueous solubility of drug discovery molecules

We investigate the use of different Machine Learning methods to construct models for aqueous solubility. Models are based on about 4000 compounds, including an in-house set of 632 drug discovery molecules of Bayer Schering Pharma. For each method, we also consider an appropriate method to obtain error bars, in order to estimate the domain of applicability (DOA) for each model. Here, we investigate error bars from a Bayesian model (Gaussian Process (GP)), an ensemble based approach (Random Forest), and approaches based on the Mahalanobis distance to training data (for Support Vector Machine and Ridge Regression models). We evaluate all approaches in terms of their prediction accuracy (in cross-validation, and on an external validation set of 536 molecules) and in how far the individual error bars can faithfully represent the actual prediction error.     

Estimating the domain of applicability for machine learning QSAR models: a study on aqueous solubility of drug discovery molecules

We investigate the use of different Machine Learning methods to construct models for aqueous solubility. Models are based on about 4000 compounds, including an in-house set of 632 drug discovery molecules of Bayer Schering Pharma. For each method, we also consider an appropriate method to obtain error bars, in order to estimate the domain of applicability (DOA) for each model. Here, we investigate error bars from a Bayesian model (Gaussian Process (GP)), an ensemble based approach (Random Forest), and approaches based on the Mahalanobis distance to training data (for Support Vector Machine and Ridge Regression models). We evaluate all approaches in terms of their prediction accuracy (in cross-validation, and on an external validation set of 536 molecules) and in how far the individual error bars can faithfully represent the actual prediction error.     

Extraction of norflurazon residues in cotton seeds with supercritical CO2

Summary A new method for the extraction of Norflurazon residues in cotton seeds using supercritical CO₂ as the extracting fluid is described. The supercritical fluid extraction results were compared with those of the classical procedure using liquid extraction. All SFE experiments were performed using a home-made system. The method presented, besides being faster and more economical than existing methods, showed better recovery, with higher selectivity for Norflurazon extraction.     

On the suitability of the virial equation for modeling the solubility of solids in supercritical fluids

Five model systems, the van der Waals fluid, the Soave-Redlich-Kwong fluid, the Peng-Robinson fluid, the hard-sphere fluid, and the square-well fluid, are used to examine the performance of the truncated virial expansion in describing the fugacity of a solute at infinite dilution in a solvent. It is demonstrated that the virial fugacity results deteriorate at significantly lower densities as the solute becomes larger. This has consequences for attempts to describe the solubility of solids in supercritical fluids, where the virial expansion, truncated after the third virial coefficient, has been considered as a modeling option. The results of this work suggest that, for the densities and solute-to-solvent size ratios commonly encountered in supercritical extraction, the truncated virial expansion should not be expected to describe correctly the solute fugacity, and therefore any success it has in fitting solubility data should be viewed with caution.     

Measurement of the solubility of metal complexes in supercritical carbon dioxide using a UV–vis spectrometer

A new apparatus based on the circulation method was developed to measure the solubility of metal complexes in supercritical carbon dioxide (scCO₂) at a wide range of temperatures and pressures. A UV–vis spectrometer, which was connected to a small saturation cell through optical fibers, was used to determine solubility. The solubilities of cobalt(III) acetylacetonate (Co(acac)₃) and chromium(III) acetylacetonate (Cr(acac)₃) in scCO₂ were measured to check the validity of both the apparatus and the method and to accumulate new solubility data. The solubility data for Cr(acac)₃ obtained in this study were in good agreement with the data reported in the literature.The measured solubilities of Co(acac)₃ and Cr(acac)₃ were also correlated with the empirical equation including the three adjustable parameters, based on the equation proposed by Chrastil. The parameters were determined by fitting the equation to the experimental data for each metal complex and the calculated results closely replicated the experimental data.     

Evaluation of attraction terms in equations of state on the prediction of solubility of some biomolecules in supercritical carbon dioxide

In the present work,effect of theattr action terms of four recently modified Peng-Robinson(MPR)equations of state on the prediction of solubility of caffeine,cholesterol,uracil and erythromycin was studied.The attraction terms of two of these equations are linear relative to the acentric factor and for the other two are exponential.It is found that the later show less deviation.Also interaction parameters for the studied systems are obtained and the percentage of average absolute relative deviation(%AARD)in each calculation is displayed.     

Theoretical and experimental study on Chloroquine drug solubility in supercritical carbon dioxide via the thermodynamic,multi-layer perceptron neural network (MLPNN), and molecular modeling

The design and development of supercritical carbon dioxide (sc-CO₂) based processes for production of pharmaceutical micro/nanoparticles is one of the interesting research topics of pharmaceutical industries owing to its attractive advantages. The solubility of drugs in sc-CO₂ at different temperatures and pressures is an essential parameter which should be determined for this purpose. Chloroquine as a traditional antirheumatic and antimalarial agent is approved as an effective drug for the treatment of Covid-19. Pishnamazi et al. (2021) measured the solubility of this drug in sc-CO₂ at the pressure range of 120–400 bar and temperature range of 308–338 K, and correlated the obtained data using some empirical models. In this work, a comprehensive computational approach was developed to more accurately study the supercritical solubility of Chloroquine. The thermodynamic models include two equations-of-state based models (Peng-Robinson and Soave-Redlich-Kowang) and two activity coefficient-based models (modified Wilson's and UNIQUAC)), as well as, a multi-layer perceptron neural network (MLPNN)) were used for this purpose. Also, molecular modeling was performed to study the electronic structure of Chloroquine and identify the potential centers of intermolecular interactions during the dissolution process. According to the obtained results, all of the theoretical models can predict Chloroquine solubility in sc-CO₂ with acceptable accuracy. Among these models, the MLPNN model possesses the highest precision with the lowest average absolute relative deviation (AARD%) of 1.76 % and the highest R_adj value of 0.999.     

Measuring and validation for isothermal solubility data of solid 2-(3,4-Dimethoxyphenyl)-5,6,7,8-tetramethoxychromen-4-one (nobiletin) in supercritical carbon dioxide

Isothermal solubility of 2-(3,4-Dimethoxyphenyl)-5,6,7,8-tetramethoxychromen-4-one (nobiletin) in supercritical carbon dioxide at temperatures of (313, 323 and 333) K and pressures from (18 to 31) MPa was measured using an analytic-recirculation methodology, with direct determination of the molar composition of the carbon dioxide-rich phase by using high performance liquid chromatography. Results indicated that the range of the measured solubility of nobiletin was from 107 · 10⁻⁶ mol · mol⁻¹ at T = 333 K and 18.35 MPa to 182 · 10⁻⁶ mol · mol⁻¹ at T = 333 K and 31.40 MPa, with a temperature crossover around 18 MPa. The validation of the experimental solubility data was carried out by using three approaches, namely, estimation of combined expanded uncertainty for each solubility data point from experimental parameters values (⩽77 · 10⁻⁶ mol · mol⁻¹); thermodynamic consistency, verified utilizing a test adapted from tools based on Gibbs–Duhem equation and solubility modelling results; and, self-consistency, proved by correlating the solubility data with a semi-empirical model as a function of temperature, pressure and pure CO₂ density.     

Quantitative analysis of polymer additives in low density polyethylene using supercritical fluid extraction/supercritical fluid chromatography

Analysis of low concentration polymer additives has been a challenging problem. The commonly used methods of analysis involve the initial extraction of polymer additives with solvents, often in a Soxhlet apparatus, followed by liquid, size exclusion, or gas chromatography. This paper describes the on-line super-critical fluid extraction (SFE)-supercritical fluid chromatographic (SFC) determination of different additives from low density polyethylene. Cryogenic collection was used as an interface between SFE and SFC to focus the extraction eluate before transfer to an analytical SFC column for quantitative analysis.     

A comparison between neural network method and semi empirical equations to predict the solubility of different compounds in supercritical carbon dioxide

Accuracy of seven semi empirical equations for the estimation of solubility of 30 different compounds in supercritical carbon dioxide has been compared with a new neural network method. To base this comparison on a fair basis, a unique set of experimental data was used for both optimization of semi empirical equations’ parameters and training, validation and testing of neural network. Results showed that neural network method with an average relative deviation of about 5.3% was more accurate than the best semi empirical equation with an average relative deviation of about 15.96% for same compounds. It was also found that the average relative deviation of semi empirical equations varies sharply among different compounds, while this quantity is less dependent on material type for neural network method.