QSAR-based solubility model for drug-like compounds

Towards the prediction of drug solubility in binary solvent mixtures at various temperatures using machine learning

Drug solubility is an important parameter in the drug development process, yet it is often tedious and challenging to measure, especially for expensive drugs or those available in small quantities. To alleviate these challenges, machine learning (ML) has been applied to predict drug solubility as an alternative approach. However, the majority of existing ML research has focused on the predictions of aqueous solubility and/or solubility at specific temperatures, which restricts the model applicability in pharmaceutical development. To bridge this gap, we compiled a dataset of 27,000 solubility datapoints, including solubility of small molecules measured in a range of binary solvent mixtures under various temperatures. Next, a panel of ML models were trained on this dataset with their hyperparameters tuned using Bayesian optimization. The resulting top-performing models, both gradient boosted decision trees (light gradient boosting machine and extreme gradient boosting), achieved mean absolute errors (MAE) of 0.33 for LogS (S in g/100 g) on the holdout set. These models were further validated through a prospective study, wherein the solubility of four drug molecules were predicted by the models and then validated with in-house solubility experiments. This prospective study demonstrated that the models accurately predicted the solubility of solutes in specific binary solvent mixtures under different temperatures, especially for drugs whose features closely align within the solutes in the dataset (MAE < 0.5 for LogS). To support future research and facilitate advancements in the field, we have made the dataset and code openly available. Scientific contribution Our research advances the state-of-the-art in predicting solubility for small molecules by leveraging ML and a uniquely comprehensive dataset. Unlike existing ML studies that predominantly focus on solubility in aqueous solvents at fixed temperatures, our work enables prediction of drug solubility in a variety of binary solvent mixtures over a broad temperature range, providing practical insights on the modeling of solubility for realistic pharmaceutical applications. These advancements along with the open access dataset and code support significant steps in the drug development process including new molecule discovery, drug analysis and formulation.

In Silico Analysis, Anticonvulsant Activity, and Toxicity Evaluation of Schisandrin B in Zebrafish Larvae and Mice

The aim of this study is to evaluate the anticonvulsant potential of schisandrin B, a main ingredient of Schisandra chinensis extracts. Schisandrin B showed anticonvulsant activity in the zebrafish larva pentylenetetrazole acute seizure assay but did not alter seizure thresholds in the intravenous pentylenetetrazole test in mice. Schisandrin B crosses the blood-brain barrier, which we confirmed in our in silico and in vivo analyses; however, the low level of its unbound fraction in the mouse brain tissue may explain the observed lack of anticonvulsant activity. Molecular docking revealed that the anticonvulsant activity of the compound in larval zebrafish might have been due to its binding to a benzodiazepine site within the GABAA receptor and/or the inhibition of the glutamate NMDA receptor. Although schisandrin B showed a beneficial anticonvulsant effect, toxicological studies revealed that it caused serious developmental impairment in zebrafish larvae, underscoring its teratogenic properties. Further detailed studies are needed to precisely identify the properties, pharmacological effects, and safety of schisandrin B.

Exploring the CSD Drug Subset: An Analysis of Lattice Energies and Constituent Intermolecular Interactions for the Crystal Structures of Pharmaceuticals

Intermolecular (synthonic) modelling is used for a statistical analysis of crystal lattice energies, together with their contributing intermolecular interactions for the crystallographic structures selected from the CCDC's Drug Subset (https://doi.org/10.1016/j.xphs.2018.12.011). Analysis of this selected subset reveal similarities in packing compared to other organic crystals in the CSD with linear relationships between molecular weight and unit cell volume, void space, and packing coefficient. Crystal lattice energy calculations converge within a 30 Å intermolecular radius characterised by a mean lattice energy of ca. -36 kcal mol^-1 with ca. 85% and 15% due to dispersive and electrostatic interactions, respectively. The distribution of the strongest synthons within the individual structures reveals an average strength of -5.79 kcal mol^-1. The diversity of chemical space within the drug molecules is in agreement with the analysis of atom types across the selected subset with phenyl groups being found to contribute the highest mean energy of -11.28 kcal mol^-1, highlighting the importance of aromatic interactions within pharmaceutical compounds. Despite an initial focus on Z' = 1 structures, this automated approach enables rapid and consistent quantitative analysis of lattice energy, synthon strength and functional group contributions, providing solid-form informatics for pharmaceutical R&D and a helpful basis for further investigations.

Molecular Screening and Toxicity Estimation of 260,000 Perfluoroalkyl and Polyfluoroalkyl Substances (PFASs) through Machine Learning

Perfluoroalkyl and polyfluoroalkyl substances (PFASs) are a class of chemicals widely used in industrial applications due to their exceptional properties and stability. However, they do not readily degrade in the environment and are linked to contamination and adverse health effects in humans and wildlife. To find alternatives for the most commonly used PFAS molecules that maintain their desirable chemical properties but are not adverse to biological lifeforms, a novel approach based upon machine learning is utilized. The machine learning model is trained on an existing set of PFAS molecules to generate over 260,000 novel PFAS molecules, which we dub PFAS-AI-Gen. Using molecular descriptors with known relationships to toxicity and industrial suitability followed by molecular docking and molecular dynamics simulations, this set of molecules is screened. In this manner, increasingly complex calculations are performed only for candidate molecules that are most likely to yield the desired properties of low binding affinity toward two selected protein receptors, the human pregnane x receptor (hPXR) and peroxisome proliferator-activated receptor γ (PPAR-γ), and high industrial suitability, defined by critical micelle concentration (CMC). The selection criteria of low binding affinity and high industrial suitability are relative to the popular PFAS alternative GenX. hPXR and PPAR-γ are selected as they are PFAS targets and facilitate a variety of functions, such as drug metabolism and glucose regulation, respectively. Through this approach, 22 promising new PFAS substitutes that may warrant experimental investigation are identified. This integrated approach of molecular screening and toxicity estimation may be applicable to other chemical classes.

Model agnostic generation of counterfactual explanations for molecules

An outstanding challenge in deep learning in chemistry is its lack of interpretability. The inability of explaining why a neural network makes a prediction is a major barrier to deployment of AI models. This not only dissuades chemists from using deep learning predictions, but also has led to neural networks learning spurious correlations that are difficult to notice. Counterfactuals are a category of explanations that provide a rationale behind a model prediction with satisfying properties like providing chemical structure insights. Yet, counterfactuals have been previously limited to specific model architectures or required reinforcement learning as a separate process. In this work, we show a universal model-agnostic approach that can explain any black-box model prediction. We demonstrate this method on random forest models, sequence models, and graph neural networks in both classification and regression.

A risk classification strategy for migrants of food contact material combined with three (Q)SAR tools in silico

The chemical constituents in food contact materials (FCMs) may transfer into food during the contact, which may pose potential risk to humans. So, it is important to evaluate the safety of FCMs. Due to the advantages of cost-effectiveness and high throughput, (Q)SAR tools have been gradually used for risk assessment. In this work, a risk classification strategy for migrants of food contact materials combined with three (Q)SAR tools was developed based on a single endpoint (Mutagenicity) assessment and risk matrix approach, respectively. 419 migrants existing in a self-built toxicology database beneficial from Python crawler technology were evaluated. 5 toxic hazard ranks and 4 risk ranks were obtained for single endpoint assessment and risk matrix respectively, with 21 substances assigned as Toxic hazard Class I and 43 substances assigned as RISK Ⅰ which need the highest safety concern. Besides, for the Toxic hazard Class I substances assessed by the single endpoint, 19 of them were confirmed experimentally, and all of them were overlapped in the RISK Ⅰ substances, which suggests the effectiveness and reliability of this strategy.

Pharmacophore-based drug design of AChE and BChE dual inhibitors as potential anti-Alzheimer's disease agents

For the Alzheimer's disease (AD) with complex pathogenesis, single target drugs represent one of the most effective therapeutic strategies in clinical. However, the traditional concept of "a disease, a target" is difficult to find very effective drugs, and multi-target drugs have already become new hot spot in drug development for this disease. In our present study, our efforts toward discovering new cholinesterase (ChE) inhibitors aided by computational methods will provide useful information as anti-AD agents in the future. The best 3D-QSAR acetylcholinesterase (AChE) and butyrylcholinesterase (BChE) inhibitors pharmacophore hypotheses Hypo1 A and Hypo1 B were generated and validated by HypoGen program in Discovery Studio 2016 based on the training set of flavonoids, and then they were used as 3D query for screening the ZINC database. Next, the hit molecules were then subjected to the ADMET and molecular docking study to prioritize the compounds. Finally, 6 compounds showed good estimated activities and promising ADMET properties. The result of best compound ZINC08751495 with AChE estimate activity (0.028), BChE estimate activity (1.55), AChE fit value (9.369), BChE fit value (8.415), AChE -CDOCKER ENERGY (30.22), BChE -CDOCKER ENERGY (33.13) has the potential for further development as a supplement to treat Alzheimer's disease.

Biomimetic Chromatographic Studies Combined with the Computational Approach to Investigate the Ability of Triterpenoid Saponins of Plant Origin to Cross the Blood-Brain Barrier

Biomimetic (non-cell based in vitro) and computational (in silico) studies are commonly used as screening tests in laboratory practice in the first stages of an experiment on biologically active compounds (potential drugs) and constitute an important step in the research on the drug design process. The main aim of this study was to evaluate the ability of triterpenoid saponins of plant origin to cross the blood-brain barrier (BBB) using both computational methods, including QSAR methodology, and biomimetic chromatographic methods, i.e., High Performance Liquid Chromatography (HPLC) with Immobilized Artificial Membrane (IAM) and cholesterol (CHOL) stationary phases, as well as Bio-partitioning Micellar Chromatography (BMC). The tested compounds were as follows: arjunic acid (Terminalia arjuna), akebia saponin D (Akebia quinata), bacoside A (Bacopa monnieri) and platycodin D (Platycodon grandiflorum). The pharmacokinetic BBB parameters calculated in silico show that three of the four substances, i.e., arjunic acid, akebia saponin D, and bacoside A exhibit similar values of brain/plasma equilibration rate expressed as logPSFubrain (the average logPSFubrain: -5.03), whereas the logPSFubrain value for platycodin D is -9.0. Platycodin D also shows the highest value of the unbound fraction in the brain obtained using the examined compounds (0.98). In these studies, it was found out for the first time that the logarithm of the analyte-micelle association constant (logKMA) calculated based on Foley's equation can describe the passage of substances through the BBB. The most similar logBB values were obtained for hydrophilic platycodin D, applying both biomimetic and computational methods. All of the obtained logBB values and physicochemical parameters of the molecule indicate that platycodin D does not cross the BBB (the average logBB: -1.681), even though the in silico estimated value of the fraction unbound in plasma is relatively high (0.52). As far as it is known, this is the first paper that shows the applicability of biomimetic chromatographic methods in predicting the penetration of triterpenoid saponins through the BBB.

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.

Multiple approaches for achieving drug solubility: an in silico perspective

Discovering new therapeutically active molecules is the ultimate destination in pharmaceutical research and development. Most drugs discovered are lipophilic and, hence, exhibit poor aqueous solubility, resulting in low bioavailability. Thus, there is a need to use various solubility enhancement techniques. Computational approaches enable the analysis of drug-carrier interactions or the numerous conformational changes in the carrier matrix that might establish an appropriate balance between cohesive and adhesive stability in a formulation. In this review, we discuss research approaches that provided molecular insight into drugs and their modifiers to unravel their solubility, stability, and bioavailability.

Entering the era of computationally driven drug development

Historically, failure rates in drug development are high; increased sophistication and investment throughout the process has shifted the reasons for attrition, but the overall success rates have remained stubbornly and consistently low. Only 8% of new entities entering clinical testing gain regulatory approval, indicating that significant obstacles still exist for efficient therapeutic development. The continued high failure rate can be partially attributed to the inability to link drug exposure with the magnitude of observed safety and efficacy-related pharmacodynamic (PD) responses; frequently, this is a result of nonclinical models exhibiting poor prediction of human outcomes across a wide range of disease conditions, resulting in faulty evaluation of drug toxicology and efficacy. However, the increasing quality and standardization of experimental methods in preclinical stages of testing has created valuable data sets within companies that can be leveraged to further improve the efficiency and accuracy of preclinical prediction for both pharmacokinetics (PK) and PD. Models of Quantitative structure-activity relationships (QSAR), physiologically based pharmacokinetics (PBPK), and PK/PD relationships have also improved efficiency. Founded on a core understanding of biochemistry and physiological interactions of xenobiotics, these in silico methods have the potential to increase the probability of compound success in clinical trials. Integration of traditional computational methods with machine-learning approaches and existing internal pharma databases stands to make a fundamental impact on the speed and accuracy of predictions during the process of drug development and approval.

Improved Prediction of Aqueous Solubility of Novel Compounds by Going Deeper With Deep Learning

Aqueous solubility is an important physicochemical property of compounds in anti-cancer drug discovery. Artificial intelligence solubility prediction tools have scored impressive performances by employing regression, machine learning, and deep learning methods. The reported performances vary significantly partly because of the different datasets used. Solubility prediction on novel compounds needs to be improved, which may be achieved by going deeper with deep learning. We constructed deeper-net models of ~20-layer modified ResNet convolutional neural network architecture, which were trained and tested with 9,943 compounds encoded by molecular fingerprints. Retrospectively tested by 62 recently-published novel compounds, one deeper-net model outperformed four established tools, shallow-net models, and four human experts. Deeper-net models also outperformed others in predicting the solubility values of a series of novel compounds newly-synthesized for anti-cancer drug discovery. Solubility prediction may be improved by going deeper with deep learning. Our deeper-net models are accessible at http://www.npbdb.net/solubility/index.jsp.

Erxian Decoction Attenuates TNF-α Induced Osteoblast Apoptosis by Modulating the Akt/Nrf2/HO-1 Signaling Pathway

Erxian decoction (EXD), a traditional Chinese medicine formula, has been used for treatment of osteoporosis for many years. The purpose of this study was to investigate the pharmacological effect of EXD in preventing osteoblast apoptosis and the underlying mechanism of prevention. Putative targets of EXD were predicted by network pharmacology, and functional and pathway enrichment analyses were also performed. Evaluations of bone mineral density, serum estradiol level, trabecular area fraction, serum calcium levels, and tumor necrosis factor (TNF)-α levels in ovariectomized rats, as well as cell proliferation assays, apoptosis assays, and western blotting in MC3T3-E1 osteoblasts were performed for further experimental validation. Ninety-three active ingredients in the EXD formula and 259 potential targets were identified. Functional and pathway enrichment analyses indicated that EXD significantly influenced the PI3K-Akt signaling pathway. In vivo experiments indicated that EXD treatment attenuated bone loss and decreased TNF-α levels in rats with osteoporosis. In vitro experiments showed that EXD treatment increased cell viability markedly and decreased levels of caspase-3 and the rate of apoptosis. It also promoted phosphorylation of Akt, nuclear translocation of transcription factor NF-erythroid 2-related factor (Nrf2), and hemeoxygenase-1 (HO-1) expression in TNF-α-induced MC3T3-E1 cells. Our results suggest that EXD exerted profound anti-osteoporosis effects, at least partially by reducing production of TNF-α and attenuating osteoblast apoptosis via Akt/Nrf2/HO-1 signaling pathway.

Strategies of Virtual Screening in Medicinal Chemistry

Virtual screening represents an effective computational strategy to rise-up the chances of finding new bioactive compounds by accelerating the time needed to move from an initial intuition to market. Classically, the most pursued approaches rely on ligand- and structure-based studies, the former employed when structural data information about the target is missing while the latter employed when X-ray/NMR solved or homology models are instead available for the target. The authors will focus on the most advanced techniques applied in this area. In particular, they will survey the key concepts of virtual screening by discussing how to properly select chemical libraries, how to make database curation, how to applying and- and structure-based techniques, how to wisely use post-processing methods. Emphasis will be also given to the most meaningful databases used in VS protocols. For the ease of discussion several examples will be presented.

Applications of Chemoinformatics in Predictive Toxicology for Regulatory Purposes, Especially in the Context of the EU REACH Legislation

Chemoinformatics methodologies such as QSAR/QSPR have been used for decades in drug discovery projects, especially for the finding of new compounds with therapeutic properties and the optimization of ADME properties on chemical series. The application of computational techniques in predictive toxicology is much more recent, and they are experiencing an increasingly interest because of the new legal requirements imposed by national and international regulations. In the pharmaceutical field, the US Food and Drug Administration (FDA) support the use of predictive models for regulatory decision-making when assessing the genotoxic and carcinogenic potential of drug impurities. In Europe, the REACH legislation promotes the use of QSAR in order to reduce the huge amount of animal testing needed to demonstrate the safety of new chemical entities subjected to registration, provided they meet specific conditions to ensure their quality and predictive power. In this review, the authors summarize the state of art of in silico methods for regulatory purposes, with especial emphasis on QSAR models.

Quantitative Regression Models for the Prediction of Chemical Properties by an Efficient Workflow

Rapid safety assessment is more and more needed for the increasing chemicals both in chemical industries and regulators around the world. The traditional experimental methods couldn't meet the current demand any more. With the development of the information technology and the growth of experimental data, in silico modeling has become a practical and rapid alternative for the assessment of chemical properties, especially for the toxicity prediction of organic chemicals. In this study, a quantitative regression workflow was built by KNIME to predict chemical properties. With this regression workflow, quantitative values of chemical properties can be obtained, which is different from the binary-classification model or multi-classification models that can only give qualitative results. To illustrate the usage of the workflow, two predictive models were constructed based on datasets of Tetrahymena pyriformis toxicity and Aqueous solubility. The qcv (2) and qtest (2) of 5-fold cross validation and external validation for both types of models were greater than 0.7, which implies that our models are robust and reliable, and the workflow is very convenient and efficient in prediction of various chemical properties.

A reliable computational workflow for the selection of optimal screening libraries

Background

The experimental screening of compound collections is a common starting point in many drug discovery projects. Successes of such screening campaigns critically depend on the quality of the screened library. Many libraries are currently available from different vendors yet the selection of the optimal screening library for a specific project is challenging. We have devised a novel workflow for the rational selection of project-specific screening libraries.

Results

The workflow accepts as input a set of virtual candidate libraries and applies the following steps to each library: (1) data curation; (2) assessment of ADME/T profile; (3) assessment of the number of promiscuous binders/frequent HTS hitters; (4) assessment of internal diversity; (5) assessment of similarity to known active compound(s) (optional); (6) assessment of similarity to in-house or otherwise accessible compound collections (optional). For ADME/T profiling, Lipinski’s and Veber’s rule-based filters were implemented and a new blood brain barrier permeation model was developed and validated (85 and 74 % success rate for training set and test set, respectively). Diversity and similarity descriptors which demonstrated best performances in terms of their ability to select either diverse or focused sets of compounds from three databases (Drug Bank, CMC and CHEMBL) were identified and used for diversity and similarity assessments. The workflow was used to analyze nine common screening libraries available from six vendors. The results of this analysis are reported for each library providing an assessment of its quality. Furthermore, a consensus approach was developed to combine the results of these analyses into a single score for selecting the optimal library under different scenarios.

Conclusions

We have devised and tested a new workflow for the rational selection of screening libraries under different scenarios. The current workflow was implemented using the Pipeline Pilot software yet due to the usage of generic components, it can be easily adapted and reproduced by computational groups interested in rational selection of screening libraries. Furthermore, the workflow could be readily modified to include additional components. This workflow has been routinely used in our laboratory for the selection of libraries in multiple projects and consistently selects libraries which are well balanced across multiple parameters.Graphical abstract

In silico Prediction of Aqueous Solubility: a Comparative Study of Local and Global Predictive Models

32 Quantitative Structure-Property Relationship (QSPR) models were constructed for prediction of aqueous intrinsic solubility of liquid and crystalline chemicals. Data sets contained 1022 liquid and 2615 crystalline compounds. Multiple Linear Regression (MLR), Support Vector Machine (SVM) and Random Forest (RF) methods were used to construct global models, and k-nearest neighbour (kNN), Arithmetic Mean Property (AMP) and Local Regression Property (LoReP) were used to construct local models. A set of the best QSPR models was obtained: for liquid chemicals with RMSE (root mean square error) of prediction in the range 0.50-0.60 log unit; for crystalline chemicals 0.80-0.90 log unit. In the case of global models the large number of descriptors makes mechanistic interpretation difficult. The local models use only one or two descriptors, so that a medicinal chemist working with sets of structurally-related chemicals can readily estimate their solubility. However, construction of stable local models requires the presence of closely related neighbours for each chemical considered. It is probable that a consensus of global and local QSPR models will be the optimal approach for construction of stable predictive QSPR models with mechanistic interpretation.

Comparing multilabel classification methods for provisional biopharmaceutics class prediction

The biopharmaceutical classification system (BCS) is now well established and utilized for the development and biowaivers of immediate oral dosage forms. The prediction of BCS class can be carried out using multilabel classification. Unlike single label classification, multilabel classification methods predict more than one class label at the same time. This paper compares two multilabel methods, binary relevance and classifier chain, for provisional BCS class prediction. Large data sets of permeability and solubility of drug and drug-like compounds were obtained from the literature and were used to build models using decision trees. The separate permeability and solubility models were validated, and a BCS validation set of 127 compounds where both permeability and solubility were known was used to compare the two aforementioned multilabel classification methods for provisional BCS class prediction. Overall, the results indicate that the classifier chain method, which takes into account label interactions, performed better compared to the binary relevance method. This work offers a comparison of multilabel methods and shows the potential of the classifier chain multilabel method for improved biological property predictions for use in drug discovery and development.

Hit identification of novel heparanase inhibitors by structure- and ligand-based approaches

Heparanase is a key enzyme involved in the dissemination of metastatic cancer cells. In this study a combination of in silico techniques and experimental methods was used to identify new potential inhibitors against this target. A 3D model of heparanase was built from sequence homology and applied to the virtual screening of a library composed of 27 known heparanase inhibitors and a commercial collection of drugs and drug-like compounds. The docking results from this campaign were combined with those obtained from a pharmacophore model recently published based in the same set of chemicals. Compounds were then ranked according to their theoretical binding affinity, and the top-rated commercial drugs were selected for further experimental evaluation. Biophysical methods (NMR and SPR) were applied to assess experimentally the interaction of the selected compounds with heparanase. The binding site was evaluated via competition experiments, using a known inhibitor of heparanase. Three of the selected drugs were found to bind to the active site of the protein and their KD values were determined. Among them, the antimalarial drug amodiaquine presented affinity towards the protein in the low-micromolar range, and was singled out for a SAR study based on its chemical scaffold. A subset of fourteen 4-arylaminoquinolines from a global set of 249 analogues of amodiaquine was selected based on the application of in silico models, a QSAR solubility prediction model and a chemical diversity analysis. Some of these compounds displayed binding affinities in the micromolar range.

High-Dimensional descriptor selection and computational QSAR modeling for antitumor activity of ARC-111 analogues Based on Support Vector Regression (SVR)

To design ARC-111 analogues with improved efficiency, we constructed the QSAR of 22 ARC-111 analogues with RPMI8402 tumor cells. First, the optimized support vector regression (SVR) model based on the literature descriptors and the worst descriptor elimination multi-roundly (WDEM) method had similar generalization as the artificial neural network (ANN) model for the test set. Secondly, seven and 11 more effective descriptors out of 2,923 features were selected by the high-dimensional descriptor selection nonlinearly (HDSN) and WDEM method, and the SVR models (SVR3 and SVR4) with these selected descriptors resulted in better evaluation measures and a more precise predictive power for the test set. The interpretability system of better SVR models was further established. Our analysis offers some useful parameters for designing ARC-111 analogues with enhanced antitumor activity.