Prediction Of Drug Solubility Research Articles

Predicting drug solubility in organic solvents mixtures: A machine-learning approach supported by high-throughput experimentation

A novel approach based on supervised machine-learning is proposed to predict the solubility of drugs and drug-like molecules in mixtures of organic solvents. Similar to quantitative structure–property relationship (QSPR) models, different solvent types are identified by molecular descriptors, which, in this study, are considered as UNIFAC subgroups. To overcome the potential lack of UNIFAC subgroups for the complex Active Pharmaceutical Ingredients (APIs) currently developed in the pharmaceutical industry, the API molecule is considered as a unique entity in the proposed modelling approach. Therefore, API solubility is predicted as a function of temperature, functional subgroups of the solvents and composition of the solvent mixture; in turn, regressors’ correlation is handled through Partial Least-Squares (PLS) regression. The method is developed and tested with experimental data of a real API and 14 organic solvents that are industrially employed for crystallisation. Solubility predictions are accurate and precise for single solvents, binary mixtures and ternary mixtures of organic solvents at different compositions and temperatures, with a determination coefficient R2 ≥ 0.90.To further test the applicability of the model, the proposed approach is applied to 9 literature organic solubility datasets of drugs and drug-like compounds and compared to benchmark solubility models in the literature. Results show that the proposed approach provides satisfactory predictions: the majority of validation and calibration data have R2 = 0.95–0.99; the ratio between RMSE (root mean squared error) of the proposed method and the range of measured solubility values is from 1 to 3 orders of magnitude smaller than the RMSE ratio obtained by the benchmark models.

Using Molecular Conformers in COSMO-RS to Predict Drug Solubility in Mixed Solvents

Using Molecular Conformers in COSMO-RS to Predict Drug Solubility in Mixed Solvents

A novel strategy of coupling artificial intelligence with chemical fingerprinting to predict drug phase behaviors in complex systems

AbstractWith the large‐scale development of drugs, understanding the drug phase behaviors in complex systems becomes more and more important. Among them, the solubility of drugs in biorelevant media needs to be urgently understood. To address this challenge, new strategies based on machine learning models are proposed. First, the strategy trains five machine learning models (extra trees [ET], gradient boosting [GB], k‐nearest neighbors [k‐NN], random forest [RF], and extreme gradient boosting [XGBoost]) based on 15 molecular descriptors of the drug molecular properties. The XGboost model was identified as the best predictive model for predicting drug solubility performance in various solvents. Next, the input feature vectors were expanded for machine learning using the MACCS chemical fingerprint coupled with the XGboost model. The MACCS chemical fingerprint coupled with XGboost model has significantly improved the prediction accuracy of drug solubility. This finding demonstrates that the proposed strategy has solubility prediction capability, which is expected to provide valid information for drug development and drug solvent screening.

Predicting Drug Solubility Using Different Machine Learning Methods - Linear Regression Model with Extracted Chemical Features vs Graph Convolutional Neural Network

Predicting Drug Solubility Using Different Machine Learning Methods - Linear Regression Model with Extracted Chemical Features vs Graph Convolutional Neural Network

Numerical optimization of drug solubility inside the supercritical carbon dioxide system using different machine learning models

This research comprehensively investigates the solubility characteristics of five distinct drugs including: Nystatin, Niflumic acid, Tolfenamic acid, Glibenclamide, and Rivaroxaban, across a range of pressure (P) and temperature (T) conditions. The solubility is computed in supercritical carbon dioxide as the solvent. It was aimed to build a holistic view of solubility estimation using machine learning technique. To predict drug solubility accurately, three regression models— K-Nearest Neighbors (KNN), Multilayer Perceptron (MLP), and Polynomial Regression (PR)—were employed, with hyperparameter optimization conducted using the Harmony Search (HS) algorithm. Performance evaluation metrics, including R-squared (R2) scores, Root Mean Square Error (RMSE), and Maximum Error, were employed to assess model effectiveness. Notably, HS-PR emerged as the top-performing model, achieving an impressive score of 0.96449 in terms of R2 metric, highlighting its proficiency in modeling drug solubility under varying conditions.

An advanced computational method for studying drug nanonization using green supercritical-based processing for improvement of pharmaceutical bioavailability in aqueous media

In this study, we implemented and compared various non-mechanistic based models for prediction of drug solubility in supercritical solvent. The data were collected from references and the models were built considering various operational circumstances. Small data sets, like the solubility data used in this study, have always been one of the challenges for modeling in machine learning method. In this study, in order to solve the regression problem related to the solubility of drugs, which includes 32 laboratory data, we implemented and studied models that are naturally compatible with very small data like solubility data of drugs in solvents. These models included Random Forest (RF), KNN and Extra Tree (ET). After obtaining the best settings for each model, their final results were compared in terms of accuracy for predicting drug solubility. The ET model had the best result with a score of 0.9999 on the R2 criterion. Random forests with 0.978 and KNN with 0.972 also had acceptable regression results. Finally, the trained model was used to display and evaluate the effect of input parameters like pressure and temperature on drug solubility to understand the process.

Theoretical investigations on the manufacture of drug nanoparticles using green supercritical processing: Estimation and prediction of drug solubility in the solvent using advanced methods

We studied the possibility of development of a novel green methodology for preparation of nanomedicine. The designed process does not use organic solvent in the manufacturing the drug and supercritical gas is used as solvent in the process. The method is efficient to enhance the drug efficacy for the patients. We used a dataset of solubility data in our study, which includes two inputs, pressure, and temperature, as well as one output, solubility, in order to carry out the study. In fact, the solubility of a drug namely Tolmetin has been predicted in supercritical carbon dioxide as the solvent using machine learning based models. As part of this research, a heap-based optimizer (HBO) was used on three selected machine learning models to obtain optimal estimators that can be used in the future. Machine learning models are multilayer perceptron (MLP), polynomial support vector regression (PSVR), and ridge regression. PSVR, Ridge, and MLP each have R2-scores of 0.976, 0.749, and 0.957, and MSEs of 1.81 × 10−8, 1.40 × 10−7, and 4.18 × 10−8, respectively. So, PSVR was selected as the most accurate model among all models assessed in this work for description of Tolmetin solubility in supercritical CO2. The results indicated that the developed models are robust and accurate enough for prediction of the pharmaceutical solubility in different solvents and operational ranges.

Development of GBRT Model as a Novel and Robust Mathematical Model to Predict and Optimize the Solubility of Decitabine as an Anti-Cancer Drug.

The efficient production of solid-dosage oral formulations using eco-friendly supercritical solvents is known as a breakthrough technology towards developing cost-effective therapeutic drugs. Drug solubility is a significant parameter which must be measured before designing the process. Decitabine belongs to the antimetabolite class of chemotherapy agents applied for the treatment of patients with myelodysplastic syndrome (MDS). In recent years, the prediction of drug solubility by applying mathematical models through artificial intelligence (AI) has become known as an interesting topic due to the high cost of experimental investigations. The purpose of this study is to develop various machine-learning-based models to estimate the optimum solubility of the anti-cancer drug decitabine, to evaluate the effects of pressure and temperature on it. To make models on a small dataset in this research, we used three ensemble methods, Random Forest (RFR), Extra Tree (ETR), and Gradient Boosted Regression Trees (GBRT). Different configurations were tested, and optimal hyper-parameters were found. Then, the final models were assessed using standard metrics. RFR, ETR, and GBRT had R2 scores of 0.925, 0.999, and 0.999, respectively. Furthermore, the MAPE metric error rates were 1.423 × 10−1 7.573 × 10−2, and 7.119 × 10−2, respectively. According to these facts, GBRT was considered as the primary model in this paper. Using this method, the optimal amounts are calculated as: P = 380.88 bar, T = 333.01 K, Y = 0.001073.

ADME prediction with KNIME: A retrospective contribution to the second "Solubility Challenge".

Computational models for predicting aqueous solubility from the molecular structure represent a promising strategy from the perspective of drug design and discovery. Since the first “Solubility Challenge”, these initiatives have marked the state-of-art of the modelling algorithms used to predict drug solubility. In this regard, the quality of the input experimental data and its influence on model performance has been frequently discussed. In our previous study, we developed a computational model for aqueous solubility based on recursive random forest approaches. The aim of the current commentary is to analyse the performance of this already trained predictive model on the molecules of the second “Solubility Challenge”. Even when our training set has inconsistencies related to the pH, solid form and temperature conditions of the solubility measurements, the model was able to predict the two sets from the second “Solubility Challenge” with statistics comparable to those of the top ranked models. Finally, we provided a KNIME automated workflow to predict aqueous solubility of new drug candidates, during the early stages of drug discovery and development, for ensuring the applicability and reproducibility of our model.

Novel Computational Approach by Combining Machine Learning with Molecular Thermodynamics for Predicting Drug Solubility in Solvents

Novel Computational Approach by Combining Machine Learning with Molecular Thermodynamics for Predicting Drug Solubility in Solvents

Fasted and fed state human duodenal fluids: Characterization, drug solubility, and comparison to simulated fluids and with human bioavailability

Accurate in vivo predictions of intestinal absorption of low solubility drugs require knowing their solubility in physiologically relevant dissolution media. Aspirated human intestinal fluids (HIF) are the gold standard, followed by simulated intestinal HIF in the fasted and fed state (FaSSIF/FeSSIF). However, current HIF characterization data vary, and there is also some controversy regarding the accuracy of FaSSIF and FeSSIF for predicting drug solubility in HIF. This study aimed at characterizing fasted and fed state duodenal HIF from 16 human volunteers with respect to pH, buffer capacity, osmolarity, surface tension, as well as protein, phospholipid, and bile salt content. The fasted and fed state HIF samples were further used to investigate the equilibrium solubility of 17 representative low-solubility small-molecule drugs, six of which were confidential industry compounds and 11 were known and characterized regarding chemical diversity. These solubility values were then compared to reported solubility values in fasted and fed state HIF, FaSSIF and FeSSIF, as well as with their human bioavailability for both states. The HIF compositions corresponded well to previously reported values and current FaSSIF and FeSSIF compositions. The drug solubility values in HIF (both fasted and fed states) were also well in line with reported solubility data for HIF, as well as simulated FaSSIF and FeSSIF. This indicates that the in vivo conditions in the proximal small intestine are well represented by simulated intestinal fluids in both composition and drug equilibrium solubility. However, increased drug solubility in the fed vs. fasted states in HIF did not correlate with the human bioavailability changes of the same drugs following oral administration in either state.

Prediction of drug solubility in ethylene glycol + water mixtures using generally trained cosolvency models

This work reports the results obtained from the generally trained versions of Yalkowsky and Jouyban-Acree models for prediction of drugs' solubility in binary mixtures of ethylene glycol + water at different temperatures. To achieve a full predictive model, the Abraham solvation solute parameters are combined with the Jouyban-Acree and Jouyban-Acree-van't Hoff models. In this respect, the solubility data sets of 14 drugs with various polarities and structural features in aqueous ethylene glycol mixtures are collected from literature and correlated with the above-mentioned models, the overall mean relative deviation values for the back-calculated drug's solubility indicating the generally trained models have an acceptable performance in solubility prediction and can be useful in pharmaceutical industry.

Pushing the limits of solubility prediction via quality-oriented data selection.

SummaryAccurate prediction of the solubility of chemical substances in solvents remains a challenge. The sparsity of high-quality solubility data is recognized as the biggest hurdle in the development of robust data-driven methods for practical use. Nonetheless, the effects of the quality and quantity of data on aqueous solubility predictions have not yet been scrutinized. In this study, the roles of the size and the quality of data sets on the performances of the solubility prediction models are unraveled, and the concepts of actual and observed performances are introduced. In an effort to curtail the gap between actual and observed performances, a quality-oriented data selection method, which evaluates the quality of data and extracts the most accurate part of it through statistical validation, is designed. Applying this method on the largest publicly available solubility database and using a consensus machine learning approach, a top-performing solubility prediction model is achieved.

Solubility of paracetamol in the ternary solvent mixtures of water + ethanol + glycerol at 298.2 and 303.2 K

ABSTRACT The solubility of paracetamol in the ternary mixtures of water + ethanol + glycerol at 298.2 and 303.2 K is determined via the shake-flask method. The reported data extend the solubility database for paracetamol in various solvent mixtures and also is employed to investigate the prediction capability of the cosolvency models for solubility prediction in the sub-binary and the ternary solvent systems. The solubility data are correlated with some cosolvency models and their accuracies are evaluated by the mean relative deviation (MRD) computed for back-calculated solubilities. The results show that the investigated models fit well and can be applied to the ternary solvent system, and the experimental data is relatively more consistent with the calculated values. The prediction ability of the trained Jouyban-Acree model for predicting drug solubility behaviour in the corresponding sub-binary systems is also investigated.

Gastrointestinal diseases and their impact on drug solubility: Celiac disease

The aim of this study was to develop an in vitro tool for predicting drug solubility and dissolution in intestinal fluids of patients with Celiac disease (CED). Biorelevant media for patients with CED were developed based on published information and a Design of Experiment (DoE) approach. The CED biorelevant media were characterised according to their surface tension, osmolality, dynamic viscosity and buffer capacity. By performing solubility studies of six drugs with different physicochemical properties in CED media, we aimed to identify drugs at high risk of altered luminal solubility in CED patients. Identified differences in CED patients compared to healthy subjects were related to a higher concentration of bile salts, lecithin and cholesterol and included as factors in the DoE resulting in 8 CED biorelevant media. Differences in media properties were observed for the surface tension between biorelevant media based on CED patients and healthy subjects. In terms of solubility, only a minimal effect of CED on the solubility of the hydrophilic neutral compound azathioprine was observed. For neutral moderately lipophilic compounds (budesonide, celecoxib), a higher surfactant concentration resulted in most cases in a higher drug solubility, while it was specific to each drug whether this was mainly driven by bile salts or lecithin. In comparison, drug solubilisation of ionisable compounds with moderate to high lipophilicity was less impacted by CED differences. The developed biorelevant CED media serve as in vitro tool to identify the main media factors impacting on drug solubility.

Machine Learning Derived Quantitative Structure Property Relationship (QSPR) to Predict Drug Solubility in Binary Solvent Systems

Prediction of drug solubility is a crucial problem in pharmaceutical industries for both drug delivery and discovery purposes. Several theoretical approaches have been proposed to predict drug solubility in mixed solvent systems when the solubility values in pure solvents are known. Quantitative structure property relationship (QSPR) approaches are gaining attention to predict various physical properties due to their robustness and computational tractability. In this work, a machine learning based QSPR approach is proposed to predict drug solubility in binary solvent systems using structural features, such as molar refractivity, McGowan volume, topological surface area, and so forth. A genetic algorithm based feature selection procedure is used to check the dependency between the selected features and to obtain the final set of significant features. Initially, solubility is assumed to behave linearly with respect to the structural features and model parameters are estimated using ordinary least-squares an...

Partial Solvation Parameters of Drugs as a New Thermodynamic Tool for Pharmaceutics.

Partial solvation parameters (PSP) have much in common with the Hansen solubility parameter or with a linear solvation energy relationship (LSER), but there are advantages based on the sound thermodynamic basis. It is, therefore, surprising that PSP has so far not been harnessed in pharmaceutics for the selection of excipients or property estimation of formulations and their components. This work introduces PSP calculation for drugs, where the raw data were obtained from inverse gas chromatography. It was shown that only a few probe gases were needed to get reasonable estimates of the drug PSPs. Interestingly, an alternative calculation of LSER parameters in silico did not reflect the experimentally obtained activity coefficients for all probe gases as well, which was attributed to the complexity of the drug structures. The experimental PSPs were proven to be helpful in predicting drug solubility in various solvents and the PSP framework allowed calculation of the different surface energy contributions. A specific benefit of PSP is that parameters can be readily converted to either classical solubility or LSER parameters. Therefore, PSP is not just about a new definition of solvatochromic parameters, but the underlying thermodynamics provides a unified approach, which holds much promise for broad applications in pharmaceutics.

Prediction of drug solubility on parallel computing architecture by support vector machines

Recent great advances in the field of high-performance computing offer new opportunities in bioinformatics, computational chemistry, and computational biology. In this paper, we use the computational methods, for instance, the support vector machines (SVM) to optimize the prediction of solubility of compounds. SVMs are trained with known data of soluble and insoluble compounds of a database, and such information is subsequently used to improve the prediction obtained by virtual screening technique. The use of larger databases increases the probability generating leads or hits, the necessary calculation time increases with the volume of the database and the accuracy of virtual screening methods. We discuss the benefits of the use of massively parallel architectures, in particular, the graphics processing units. We empirically demonstrate that the graphics processing units are well adapted for the methodical acceleration of the support vector machines, of the order up to 45 times, compared to their sequential version.

Generally trained models to predict drug solubility in methanol + water mixtures

The aim of this study is to develop the trained versions of Yalkowsky and Jouyban-Acree models for the prediction of drug solubility in the binary aqueous mixtures of methanol (MeOH) at various temperatures. To provide a full predictive model, the Abraham solvation parameters of solutes are combined with the proposed models. The solubility data of 41 drug and/or drug-like compounds with different polarities and structural features covering the total drug-like space are fitted by these models. The generally trained models provide reasonable estimation of the solubility behavior of drugs and can be helpful in the pharmaceutical industry.

Generally trained models to predict drug solubility in N-methyl-2-pyrrolidone + water mixtures at various temperatures

The trained versions of Yalkowsky and Jouyban-Acree models are proposed to predict solubility of drugs in binary aqueous mixtures of N-methyl-2-pyrrolidone (NMP) at various temperatures. To provide a full predictive model, the Abraham solvation parameters of solutes are combined with the Jouyban-Acree model and Jouyban-Acree-van't Hoff model. Since this investigation includes a number of compounds with different polarities and structural features, the results may provide accurate estimations of solubilization for most compounds of interest. The overall mean relative deviation (MRD) values for the back-calculated solubility of drugs in {NMP + water} solvent mixtures are 20.9–39.3% for the proposed models indicating that the generally trained models provided acceptable predictions and could be helpful in the pharmaceutical industry.