Prediction Of Aqueous Solubility Research Articles

Creation and interpretation of machine learning models for aqueous solubility prediction

Aim: Solubility prediction is an essential factor in rational drug design and many models have been developed with machine learning (ML) methods to enhance the predictive ability. However, most of the ML models are hard to interpret which limits the insights they can give in the lead optimization process. Here, an approach to construct and interpret solubility models with a combination of physicochemical properties and ML algorithms is presented. Methods: The models were trained, optimized, and tested in a dataset containing 12,983 compounds from two public datasets and further evaluated in two external test sets. More importantly, the SHapley Additive exPlanations (SHAP) and heat map coloring approaches were used to explain the predictive models and assess their suitability to guide compound optimization. Results: Among the different ML methods, random forest (RF) models obtain the best performance in the different test sets. From the interpretability perspective, fragment-based coloring offers a more robust interpretation than atom-based coloring and that normalizing the values further improves it. Conclusions: Overall, for certain applications simple ML algorithms such as RF work well and can outperform more complex methods and that combining them with fragment-coloring can offer guidance for chemists to modify the structure with a desired property. This interpretation strategy is publicly available at https://github.com/Pharmacelera/predictive-model-coloring and could be further applied in other property predictions to improve the interpretability of ML models.

Prediction of organic compound aqueous solubility using machine learning: a comparison study of descriptor-based and fingerprints-based models

A reliable and practical determination of a chemical species’ solubility in water continues to be examined using empirical observations and exhaustive experimental studies alone. Predictions of chemical solubility in water using data-driven algorithms can allow us to create a rationally designed, efficient, and cost-effective tool for next-generation materials and chemical formulations. We present results from two machine learning (ML) modeling studies to adequately predict various species’ solubility using data for over 8400 compounds. Molecular-descriptors, the most used method in previous studies, and Morgan fingerprint, a circular-based hash of the molecules' structures, were applied to produce water solubility estimates. We trained all models on 80% of the total datasets using the Random Forest (RFs) technique as the regressor and tested the prediction performance using the remaining 20%, resulting in coefficient of determination (R2) test values of 0.88 and 0.81 and root-mean-square deviation (RMSE) test values 0.64 and 0.80 for the descriptors and circular fingerprint methods, respectively. We interpreted the produced ML models and reported the most effective features for aqueous solubility measures using the Shapley Additive exPlanations (SHAP) and thermodynamic analysis. Low error, ability to investigate the molecular-level interactions, and compatibility with thermodynamic quantities made the fingerprint method a distinct model compared to other available computational tools. However, it is worth emphasizing that physicochemical descriptor model outperformed the fingerprint model in achieving better predictive accuracy for the given test set.

Revised descriptors for polycyclic aromatic and related hydrocarbons for the prediction of environmental properties using the solvation parameter model

Revised descriptors for twenty-five polycyclic aromatic and related hydrocarbons (PAHs) forming a component of the Wayne State University (WSU) descriptor database are provided for use with the solvation parameter model. The descriptors are determined by the Solver method using experimental data for calibrated gas-liquid and reversed-phase liquid chromatographic retention factors and liquid-liquid partition constants in totally organic biphasic systems. The characteristic solvation properties of the PAHs are accounted for mainly by the additional dispersion interactions (E descriptor) and dipole-type interactions (S descriptor) resulting from the availability of easily polarizable electrons that complement typical dispersion interactions for saturated hydrocarbons. The descriptors afford acceptable prediction of the water-air partition constant (average absolute deviation AAD = 0.17, n = 22), octanol-air partition constant (AAD = 0.12, n = 20), and water-octanol partition constant (AAD = 0.10, n = 23). A two-parameter model containing only the V and B descriptors provides an unbiased prediction of aqueous solubility for the PAHs with an AAD = 0.26 (n = 22). The descriptors estimated by convenient chromatographic and partition constant measurements are demonstrated to be a viable alternative to the experimental determination of environmental properties otherwise only available by tedious, expensive, and low data throughput experimental techniques.

Experimental and Virtual Screening of Phenazine and Phenothiazine Derivatives in Aqueous Solution for Electrochemical Carbon Capture

Molecular redox-active compounds, including quinone and aza-aromatics, have emerged as tools for capturing carbon dioxide using renewable electricity. Given the vast and highly diverse chemical space of the candidate compounds, accessing their electrochemical properties in a rapid way is attractive for both experimental and virtual screening approaches.Here we present a study on the experimental screening of a series of commercial redox-active compounds. Phenazine and phenothiazine dyes have been experimentally tested for applications in electrochemical carbon capture, for which they are finding increasing applications as redox mediators due to their unique redox properties in aqueous solution. Nicotinamide has been used as a redox-inactive solubilizing agent to increase the solubilities of general phenazine and phenothiazine derivatives for rapid experimental screening of commercial compounds without time-consuming synthetic modifications. Cyclic voltammetry in aqueous solutions under nitrogen and carbon dioxide indicated reversible behavior in the pH swing of an aqueous solution. Bulk electrolysis experiments demonstrated the chemical stability of phenazine and phenothiazine derivatives. The redox properties and the chemical stability make phenazine and phenothiazine derivatives interesting candidates for future use as redox-active materials in electrochemical carbon capture.We also present progress towards a virtual screening for the optimization of electrochemical and physicochemical properties of redox-active compounds. Based on our experimental screening of commercial phenazine and phenothiazine compounds identifying neutral red as the lead compound, a virtual screening was performed for optimization of electrochemical and physicochemical properties using machine learning. The virtual screening development steps include (1) chemical library generation using neutral red as a lead compound, (2) molecular property prediction based on quantum chemical calculations, (3) aqueous solubility prediction using machine learning, (4) data processing and database creation, and (5) experimental validation of the optimized compounds for the use of electrochemical carbon capture. Figure 1

An interpretable graph representation learning model for accurate predictions of drugs aqueous solubility

As increasingly more data science-driven approaches have been applied for compound properties predictions in the domain of drug discovery, such kinds of methods have displayed considerable accuracy compared to conventional ones. In this work, we proposed an interpretable graph learning representation model, SolubNet, for drug aqueous solubility prediction. The comprehensive evaluation demonstrated that SolubNet can successfully capture the quantitative structure-property relationship and can be interpreted with layer-wise relevance propagation (LRP) algorithm regarding how prediction values are generated from original input structures. The key advantage of SolubNet lies in the fact that it includes 3 layers of Topology Adaptive Graph Convolutional Networks which can efficiently perceive chemical local environments. SolubNet showed high performance in several tasks for drugs’ aqueous solubility prediction. LRP revealed that SolubNet can identify high and low polar regions of a given molecule, assigning them reasonable weights to predict the final solubility, in a way highly compatible with chemists’ intuition. We are confident that such a flexible yet interpretable and accurate tool will largely enhance the efficiency of drug discovery, and will even contribute to the methodology development of computational pharmaceutics.

Prediction of aqueous solubility of a strongly soluble solute from molecular simulation.

The prediction of solubilities of compounds by means of molecular simulation has been receiving increasing attention due to the key role played by solubility in countless applications. We have predicted the aqueous solubility of urea at 300K from chemical potential calculations for two urea model combinations: Özpinar/TIP3P and Hölzl/(TIP4P/2005). The methodology assumes that the intramolecular contribution of the urea molecule to the chemical potentials is identical in the crystal and in solution and, hence, cancels out. In parallel to the chemical potential calculations, we also performed direct coexistence simulations of a urea crystal slab in contact with urea-water solutions with the aim to identify upper and lower bounds to the solubility value using an independent route. The chemical potential approach yielded similar solubilities for both urea models, despite the actual chemical potential values showing a significant dependence on the force field. The predicted solubilities for the two models were 0.013-0.018 (Özpınar) and 0.008-0.012 (Hölzl) mole fraction, which are an order of magnitude lower than the experimental solubility that lies in a range of 0.125-0.216 mole fraction. The direct coexistence solubility bounds were relatively wide and did not encompass the chemical potential based solubilities, although the latter were close to the lower bound values.

Multiple machine learning algorithms assisted QSPR models for aqueous solubility: Comprehensive assessment with CRITIC-TOPSIS.

As an essential environmental property, the aqueous solubility quantifies the hydrophobicity of a compound. It could be further utilized to evaluate the ecological risk and toxicity of organic pollutants. Concerned about the proliferation of organic contaminants in water and the associated technical burden, researchers have developed QSPR models to predict aqueous solubility. However, there are no standard procedures or best practices on how to comprehensively evaluate models. Hence, the CRITIC-TOPSIS comprehensive assessment method was first-ever proposed according to a variety of statistical parameters in the environmental model research field. 39 models based on 13 ML algorithms (belonged to 4 tribes) and 3 descriptor screening methods, were developed to calculate aqueous solubility values (log Kws) for organic chemicals reliably and verify the effectiveness of the comprehensive assessment method. The evaluations were carried out for exhibiting better predictive accuracy and external competitiveness of the MLR-1, XGB-1, DNN-1, and kNN-1 models in contrast to other prediction models in each tribe. Further, XGB model based on SRM (XGB-1, C = 0.599) was selected as an optimal pathway for prediction of aqueous solubility. We hope that the proposed comprehensive evaluation approach could act as a promising tool for selecting the optimum environmental property prediction methods.

SOMAS: a platform for data-driven material discovery in redox flow battery development

Aqueous organic redox flow batteries offer an environmentally benign, tunable, and safe route to large-scale energy storage. The energy density is one of the key performance parameters of organic redox flow batteries, which critically depends on the solubility of the redox-active molecule in water. Prediction of aqueous solubility remains a challenge in chemistry. Recently, machine learning models have been developed for molecular properties prediction in chemistry and material science. The fidelity of a machine learning model critically depends on the diversity, accuracy, and abundancy of the training datasets. We build a comprehensive open access organic molecular database “Solubility of Organic Molecules in Aqueous Solution” (SOMAS) containing about 12,000 molecules that covers wider chemical and solubility regimes suitable for aqueous organic redox flow battery development efforts. In addition to experimental solubility, we also provide eight distinctive quantum descriptors including optimized geometry derived from high-throughput density functional theory calculations along with six molecular descriptors for each molecule. SOMAS builds a critical foundation for future efforts in artificial intelligence-based solubility prediction models.

RedDB, a computational database of electroactive molecules for aqueous redox flow batteries

An increasing number of electroactive compounds have recently been explored for their use in high-performance redox flow batteries for grid-scale energy storage. Given the vast and highly diverse chemical space of the candidate compounds, it is alluring to access their physicochemical properties in a speedy way. High-throughput virtual screening approaches, which use powerful combinatorial techniques for systematic enumerations of large virtual chemical libraries and respective property evaluations, are indispensable tools for an agile exploration of the designated chemical space. Herein, RedDB: a computational database that contains 31,618 molecules from two prominent classes of organic electroactive compounds, quinones and aza-aromatics, has been presented. RedDB incorporates miscellaneous physicochemical property information of the compounds that can potentially be employed as battery performance descriptors. RedDB’s development steps, including: (i) chemical library generation, (ii) molecular property prediction based on quantum chemical calculations, (iii) aqueous solubility prediction using machine learning, and (iv) data processing and database creation, have been described.

Prediction of the Aqueous Solubility of Compounds Based on Light Gradient Boosting Machines with Molecular Fingerprints and the Cuckoo Search Algorithm.

Aqueous solubility is one of the most important physicochemical properties in drug discovery. At present, the prediction of aqueous solubility of compounds is still a challenging problem. Machine learning has shown great potential in solubility prediction. Most machine learning models largely rely on the setting of hyperparameters, and their performance can be improved by setting the hyperparameters in a better way. In this paper, we used MACCS fingerprints to represent the structural features and optimized the hyperparameters of the light gradient boosting machine (LightGBM) with the cuckoo search algorithm (CS). Based on the above representation and optimization, the CS-LightGBM model was established to predict the aqueous solubility of 2446 organic compounds and the obtained prediction results were compared with those obtained with the other six different machine learning models (RF, GBDT, XGBoost, LightGBM, SVR, and BO-LightGBM). The comparison results showed that the CS-LightGBM model had a better prediction performance than the other six different models. RMSE, MAE, and R 2 of the CS-LightGBM model were, respectively, 0.7785, 0.5117, and 0.8575. In addition, this model has good scalability and can be used to solve solubility prediction problems in other fields such as solvent selection and drug screening.

Multi-channel GCN ensembled machine learning model for molecular aqueous solubility prediction on a clean dataset.

This study constructed a new aqueous solubility dataset and a solubility regression model which was ensembled by GCN and machine learning models. Aqueous solubility is a key physiochemical property of small molecules in drug discovery. In the past few decades, there have been many studies about solubility prediction. However, many of these studies have high root mean squared error (RMSE). Meanwhile, their dataset always contains salt compounds and solubility data obtained from different experimental conditions. In this paper, we constructed a clean dataset with 2609 compounds, which was small but contains only solubility records without salts at the same temperatures (25°C). Here, we applied graph convolutional neural network (GCN) to construct an aqueous solubility prediction model. To enhance the performance of the model, the molecular MACCS key fingerprints and physiochemical descriptors were also combined with the GCN model to build a multi-channel model. Additionally, the authors also built two machine learning models (support vector regression and gradient boost decision tree) and assembled them to the GCN model to improve the root mean squared error (RMSE = 0.665). Finally, comparative experiments have shown that our framework achieved the best performance on ESOL dataset (RMSEval = 0.56, RMSEtest = 0.44) and surpassed four established software on aqueous solubility prediction of new compounds.

Data-driven discovery of small electroactive molecules for energy storage in aqueous redox flow batteries

A high-throughput virtual screening (HTVS)-guided experimental study is applied for the large-scale exploration of quinone-like anolytes for aqueous redox flow batteries (ARFBs). This includes the design of a focused virtual chemical library inspired by stable molecules, quantum chemical prediction of redox properties, machine learning prediction of aqueous solubility, automated search for commercial availability on vendor databases, and electrochemical characterization of the promising compounds. Screening in a chemical space of 3257 redox pairs led to 205 predicted candidates with higher solubility and lower redox potential than that of anthraquinone-2,7-disulfonic acid (AQDS) anolyte used in ARFBs. Through electrochemical studies on commercially available compounds, the molecules that show good performance in a practical ARFB setup are identified. Among them, indigo trisulfonate [Indigo-3(SO3H)] showed higher solubility, capacity retention, and coulombic efficiency than AQDS and its predecessors. The data-driven material design approach is therefore highly effective for the future explorations of electrochemical energy storage compounds.

Dealing with a data-limited regime: Combining transfer learning and transformer attention mechanism to increase aqueous solubility prediction performance

Aqueous solubility is a key chemical property that drives various processes in chemistry and biology. Its computational prediction is challenging, as evidenced by the fact that it has been a subject of considerable interest for several decades. Recent work has explored fingerprint-based, feature-based and graph-based representations with different machine learning and deep learning methodologies. In general, many traditional methods have been proposed, but they rely heavily on the quality of the rule-based, hand-crafted features. On the other hand, limitations in the quality of aqueous solubility data become a handicap when training deep models. In this study, we have developed a novel structure-aware method for the prediction of aqueous solubility by introducing a new deep network architecture and then employing a transfer learning approach. The model was proven to be competitive, obtaining an RMSE of 0.587 during both cross-validation and a test on an independent dataset. To be more precise, the method is evaluated on molecules downloaded from the Online Chemical Database and Modeling Environment (OCHEM). Beyond aqueous solubility prediction, the strategy presented in this work may be useful for modeling any kind of (chemical or biological) properties for which there is a limited amount of data available for model training.

Improved Lipophilicity and Aqueous Solubility Prediction with Composite Graph Neural Networks.

The accurate prediction of molecular properties, such as lipophilicity and aqueous solubility, are of great importance and pose challenges in several stages of the drug discovery pipeline. Machine learning methods, such as graph-based neural networks (GNNs), have shown exceptionally good performance in predicting these properties. In this work, we introduce a novel GNN architecture, called directed edge graph isomorphism network (D-GIN). It is composed of two distinct sub-architectures (D-MPNN, GIN) and achieves an improvement in accuracy over its sub-architectures employing various learning, and featurization strategies. We argue that combining models with different key aspects help make graph neural networks deeper and simultaneously increase their predictive power. Furthermore, we address current limitations in assessment of deep-learning models, namely, comparison of single training run performance metrics, and offer a more robust solution.

SolTranNet-A Machine Learning Tool for Fast Aqueous Solubility Prediction.

While accurate prediction of aqueous solubility remains a challenge in drug discovery, machine learning (ML) approaches have become increasingly popular for this task. For instance, in the Second Challenge to Predict Aqueous Solubility (SC2), all groups utilized machine learning methods in their submissions. We present SolTranNet, a molecule attention transformer to predict aqueous solubility from a molecule's SMILES representation. Atypically, we demonstrate that larger models perform worse at this task, with SolTranNet's final architecture having 3,393 parameters while outperforming linear ML approaches. SolTranNet has a 3-fold scaffold split cross-validation root-mean-square error (RMSE) of 1.459 on AqSolDB and an RMSE of 1.711 on a withheld test set. We also demonstrate that, when used as a classifier to filter out insoluble compounds, SolTranNet achieves a sensitivity of 94.8% on the SC2 data set and is competitive with the other methods submitted to the competition. SolTranNet is distributed via pip, and its source code is available at https://github.com/gnina/SolTranNet.

Machine learning in prediction of intrinsic aqueous solubility of drug‐like compounds: Generalization, complexity, or predictive ability?

AbstractWe present a collection of publicly available intrinsic aqueous solubility data of 829 drug‐like compounds. Four different machine learning algorithms (random forests [RF], LightGBM, partial least squares, and least absolute shrinkage and selection operator [LASSO]) coupled with multistage permutation importance for feature selection and Bayesian hyperparameter optimization were used for the prediction of solubility based on chemical structural information. Our results show that LASSO yielded the best predictive ability on an external test set with a root mean square error (RMSE) (test) of 0.70 log points, an R2(test) of 0.80, and 105 features. Taking into account the number of descriptors as well, an RF model achieves the best balance between complexity and predictive ability with an RMSE(test) of 0.72 log points, an R2(test) of 0.78, and with only 17 features. On a more aggressive test set (principal component analysis [PCA]‐based split), better generalization was observed for the RF model. We propose a ranking score for choosing the best model, as test set performance is only one of the factors in creating an applicable model. The ranking score is a weighted combination of generalization, number of features, and test performance. Out of the two best learners, a consensus model was built exhibiting the best predictive ability and generalization with RMSE(test) of 0.67 log points and a R2(test) of 0.81.

Predicting aqueous solubility by QSPR modeling

The aqueous solubility is predicted here using quantitative structure property relationship (QSPR) models. In this study, we examine whether descriptors that individually yield favorable models for the prediction of the Gibbs energy of solvation and sublimation can be used in combination with octanol-water partition coefficient to produce QSPR models for the prediction of aqueous solubility. Based on this strategy, applied to seven distinct datasets, all models exhibited an R2 greater than 0.7 and Q2 greater than 0.6 for the estimation of aqueous solubility. We also determined how uncoupling the descriptors used to create QSPR models in the prediction of Gibbs energy of sublimation yielded an improved model. Model refinement using an artificial neural network applying the same descriptors generated significantly better models with improved R2 and standard deviation.

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.

Optimizing Pharmacokinetic Property Prediction Based on Integrated Datasets and a Deep Learning Approach.

Oral bioavailability (OBA)-related pharmacokinetic properties, such as aqueous solubility, lipophilicity, and intestinal membrane permeability, play a significant role in drug discovery. However, their measurement is usually costly and time-consuming. Therefore, prediction models based on diverse approaches have been established in recent decades. Computational prediction of molecular properties has become an important step in drug discovery, aiming to identify potential drug-like candidates and reduce costs. However, limitations related to dataset capacity and algorithm adaptation still place restrictions on the applicability of the related models. In this study, we considered both dataset and algorithm optimization to address the challenge of predicting OBA-related molecular properties. Benchmark datasets of aqueous solubility (log S), lipophilicity (log D), and membrane permeability measured using the Caco-2 cell line (log Papp) were constructed by merging and calibrating experimental data from diverse articles and databases. Then, a novel molecular property prediction model, called a multiembedding-based synthetic network (MESN), was generated by applying a deep learning algorithm based on the synthesis of multiple types of molecular embeddings. MESN achieves performance improvements over other state-of-the-art methods for the prediction of aqueous solubility, lipophilicity, and membrane permeability. Results were also obtained using several other algorithms and independent validation datasets as a control study. Moreover, a dimension reduction analysis (based on t-distributed stochastic neighbor embedding, t-SNE) and an atomic feature similarity analysis showed that the molecular embeddings extracted from the MESN model exhibit good clustering and diversity. Overall, considering the fundamental role of the data and the superior prediction performance of the model, we highlight the applicability of MESN on benchmark datasets for further utility in drug discovery-related molecular property prediction.

ADME prediction with KNIME: In silico aqueous solubility consensus model based on supervised recursive random forest approaches.

In-silico prediction of aqueous solubility plays an important role during the drug discovery and development processes. For many years, the limited performance of in-silico solubility models has been attributed to the lack of high-quality solubility data for pharmaceutical molecules. However, some studies suggest that the poor accuracy of solubility prediction is not related to the quality of the experimental data and that more precise methodologies (algorithms and/or set of descriptors) are required for predicting aqueous solubility for pharmaceutical molecules. In this study a large and diverse database was generated with aqueous solubility values collected from two public sources; two new recursive machine-learning approaches were developed for data cleaning and variable selection, and a consensus model based on regression and classification algorithms was created. The modeling protocol, which includes the curation of chemical and experimental data, was implemented in KNIME, with the aim of obtaining an automated workflow for the prediction of new databases. Finally, we compared several methods or models available in the literature with our consensus model, showing results comparable or even outperforming previous published models.