Prediction of skin permeability of drugs. I. Comparison with artificial membrane

In order to measure the contribution of lipid and pore (aqueous) pathways to the total skin permeation of drugs, and to establish a predictive method for the steady state permeation rate of drugs, the relationship between permeability through excised hairless rat skin and some physicochemical properties of several drugs were compared with those through polydimethylsiloxane (silicone) and poly(2-hydroxyethyl methacrylate) (pHEMA) membranes, as typical solution-diffusion and porous membranes, respectively. A linear relationship was found between the permeability coefficients of drugs for the silicone membrane and their octanol/water partition coefficients. For the pHEMA membrane, the permeability coefficients were almost constant independent of the partition coefficient. On the other hand, the skin permeation properties could be classified into two types: one involves the case of lipophilic drugs, where the permeability coefficient is correlated to the partition coefficient, similar to the silicone membrane; and the other involves hydrophilic drugs, where the permeability coefficients were almost constant, similar to pHEMA membrane. From the above results, the stratum corneum, the main barrier in skin, could be described as a membrane having two parallel permeation pathways: lipid and pore pathways. An equation for predicting the steady state permeation rate of drugs was derived based on this skin permeation model.     

Combinatorial QSAR modeling of chemical toxicants tested against Tetrahymena pyriformis

Selecting most rigorous quantitative structure-activity relationship (QSAR) approaches is of great importance in the development of robust and predictive models of chemical toxicity. To address this issue in a systematic way, we have formed an international virtual collaboratory consisting of six independent groups with shared interests in computational chemical toxicology. We have compiled an aqueous toxicity data set containing 983 unique compounds tested in the same laboratory over a decade against Tetrahymena pyriformis. A modeling set including 644 compounds was selected randomly from the original set and distributed to all groups that used their own QSAR tools for model development. The remaining 339 compounds in the original set (external set I) as well as 110 additional compounds (external set II) published recently by the same laboratory (after this computational study was already in progress) were used as two independent validation sets to assess the external predictive power of individual models. In total, our virtual collaboratory has developed 15 different types of QSAR models of aquatic toxicity for the training set. The internal prediction accuracy for the modeling set ranged from 0.76 to 0.93 as measured by the leave-one-out cross-validation correlation coefficient ( Q abs2). The prediction accuracy for the external validation sets I and II ranged from 0.71 to 0.85 (linear regression coefficient R absI2) and from 0.38 to 0.83 (linear regression coefficient R absII2), respectively. The use of an applicability domain threshold implemented in most models generally improved the external prediction accuracy but at the same time led to a decrease in chemical space coverage. Finally, several consensus models were developed by averaging the predicted aquatic toxicity for every compound using all 15 models, with or without taking into account their respective applicability domains. We find that consensus models afford higher prediction accuracy for the external validation data sets with the highest space coverage as compared to individual constituent models. Our studies prove the power of a collaborative and consensual approach to QSAR model development. The best validated models of aquatic toxicity developed by our collaboratory (both individual and consensus) can be used as reliable computational predictors of aquatic toxicity and are available from any of the participating laboratories.     

In vivo comparison of the optical clearing efficacy of optical clearing agents in human skin by quantifying permeability using optical coherence tomography

The objective of this work is to quantify and compare the optical clearing efficacy of glucose, propylene glycol, glycerol solutions through the human skin tissue in vivo by calculating permeability coefficient of three solutions. Currently, the permeability coefficient of agent in tissues was extracted from optical coherence tomography (OCT) amplitude data mainly through the OCT signal slope and the OCT amplitude methods. In this study, we report the OCT attenuation coefficient method which is a relatively novel and rarely reported methodology to measure the permeability coefficient during the optical skin clearing procedure. The permeability coefficients for 40% propylene glycol, glucose and glycerol were (2.74 ± 0.05) × 10(-6) cm s(-1), (1.78 ± 0.04) × 10(-6) cm s(-1) and (1.67 ± 0.04) × 10(-6) cm s(-1), respectively. It could be clearly seen that the permeability coefficient of the 40% propylene glycol solution is higher than that of 40% glucose solution, and the permeability coefficient of the 40% glucose solution is higher than that of the 40% glycerol solution. These indicate 40% propylene glycol solution is more effective than others in the human skin in vivo. We then compare and prove consistency of optical clearing efficacy figured out by three different methods.     

Structural characterization and prediction of Kovats retention indices (RI) for alkylbenzene compounds

A new molecular structural characterization (MSC) method called the molecular vertex eigenvalue correlative index (MVECI) is constructed and used to describe the structures of 122 alkylbenzene compounds. Through multiple linear regression (MLR) and stepwise multiple regression (SMR), a quantitative structure-retention relationship (QSRR) model with correlation coefficient (R) of 0.995 is obtained. Through partial least-square regression (PLS), another QSRR model with correlation coefficient (R) of 0.991 is obtained. The estimation stability and prediction ability of the two models are strictly analyzed by both internal and external validations. For the internal validation, the cross-validation (CV) correlation coefficients (R _CV) of the two models are 0.993 and 0.988. For the external validation, the correlation coefficients (R _test) of the two models are 0.996 and 0.995, respectively. The results show that the stability and predictability of the models are good, and the molecular vertex eigenvalue correlative index can successfully describe the structures of alkylbenzene compounds.     

A Recurrent Neural Network model to predict blood–brain barrier permeability

The rapid development of computational methods and the increasing volume of chemical and biological data have contributed to an immense growth in chemical research. This field of study is known as “chemoinformatics,” which is a discipline that uses machine-learning techniques to extract, process, and extrapolate data from chemical structures. One of the significant lines of research in chemoinformatics is the study of blood–brain barrier (BBB) permeability, which aims to identify drug penetration into the central nervous system (CNS). In this research, we attempt to solve the problem of BBB permeability by predicting compounds penetration to the CNS. To accomplish this goal: (i) First, an overview is provided to the field of chemoinformatics, its definition, applications, and challenges, (ii) Second, a broad view is taken to investigate previous machine-learning and deep-learning computational models to solve BBB permeability. Based on the analysis of previous models, three main challenges that collectively affect the classifier performance are identified, which we define as “the triple constraints”; subsequently, we map each constraint to a proposed solution, (iii) Finally, we conclude this endeavor by proposing a deep learning based Recurrent Neural Network model, to predict BBB permeability (RNN-BBB model). Our model outperformed other studies from the literature by scoring an overall accuracy of 96.53%, and a specificity score of 98.08%. The obtained results confirm that addressing the triple constraints substantially improves the classification model capability specifically when predicting compounds with low penetration.     

LINGO, an efficient holographic text based method to calculate biophysical properties and intermolecular similarities

SMILES strings are the most compact text based molecular representations. Implicitly they contain the information needed to compute all kinds of molecular structures and, thus, molecular properties derived from these structures. We show that this implicit information can be accessed directly at SMILES string level without the need to apply explicit time-consuming conversion of the SMILES strings into molecular graphs or 3D structures with subsequent 2D or 3D QSPR calculations. Our method is based on the fragmentation of SMILES strings into overlapping substrings of a defined size that we call LINGOs. The integral set of LINGOs derived from a given SMILES string, the LINGO profile, is a hologram of the SMILES representation of the molecule described. LINGO profiles provide input for QSPR models and the calculation of intermolecular similarities at very low computational cost. The octanol/water partition coefficient (LlogP) QSPR model achieved a correlation coefficient R2=0.93, a root-mean-square error RRMS=0.49 log units, a goodness of prediction correlation coefficient Q2=0.89 and a QRMS=0.61 log units. The intrinsic aqueous solubility (LlogS) QSPR model achieved correlation coefficient values of R2=0.91, Q2=0.82, and RRMS=0.60 and QRMS=0.89 log units. Integral Tanimoto coefficients computed from LINGO profiles provided sharp discrimination between random and bioisoster pairs extracted from Accelrys Bioster Database. Average similarities (LINGOsim) were 0.07 for the random pairs and 0.36 for the bioisosteric pairs.     

Permeability of ionized salicylate derivatives through guinea pig dorsal skin

pH-dependency of skin permeability to salicylic acid was examined in excised guinea pig dorsal skin. Permeation followed the pH-partition theory at acidic pH. However, above pH 5.0 the observed permeability coefficients were larger than the estimated values obtained from the ratio of the undissociated forms. These findings are quite different from those obtained using the same drug and a silicone rubber membrane, in which permeability coefficients were consistent with the pH-partition theory. The findings suggested that permeation of salicylate as anions occurred at a neutral skin pH. The permeability coefficient of the ionized form was estimated to be about 1.6% of the nonionized form. We also examined the skin permeability of salicylate and its five 5-substituents and two 3-substituents at pH 7.4. We investigated the relationship between their permeability coefficients and the physico-chemical properties of the substituents. Multi regression analysis on the permeability coefficients showed a parabolic relationship between the values of the hydrophobic parameter (pi) and the logarithms of the permeability coefficients. These findings suggested that the ionic permeation pathway of salicylate derivatives is controlled by hydrophobic as well as hydrophilic properties.     

Predicting skin permeability from complex chemical mixtures: incorporation of an expanded QSAR model

Quantitative structure–activity relationship (QSAR) models have been widely used to study the permeability of chemicals or solutes through skin. Among the various QSAR models, Abraham’s linear free-energy relationship (LFER) model is often employed. However, when the experimental conditions are complex, it is not always appropriate to use Abraham’s LFER model with a single set of regression coefficients. In this paper, we propose an expanded model in which one set of partial slopes is defined for each experimental condition, where conditions are defined according to solvent: water, synthetic oil, semi-synthetic oil, or soluble oil. This model not only accounts for experimental conditions but also improves the ability to conduct rigorous hypothesis testing. To more adequately evaluate the predictive power of the QSAR model, we modified the usual leave-one-out internal validation strategy to employ a leave-one-solute-out strategy and accordingly adjust the Q² _LOO statistic. Skin permeability was shown to have the rank order: water > synthetic > semi-synthetic > soluble oil. In addition, fitted relationships between permeability and solute characteristics differ according to solvents. We demonstrated that the expanded model (r² = 0.70) improved both the model fit and the predictive power when compared with the simple model (r² = 0.21).     

Pore structure of gel chitosan membranes. III. Pressuredriven mass transport measurements

The capillary pore model of water-swollen gels was used to interpret pressure-driven mass transport properties of gel chitosan membranes. Pure water hydraulic permeability coefficients, L_p, and rejection coefficients, R, of 13 solutes ranging in molecular radius from 2.4 Å (methanol) to 16 Å (polyethylene glycol 6000) were measured for an untreated chitosan membrane, for two chitosan membranes crosslinked with glutaraldehyde of concentrations 0.01 and 0.1% and coated with a protein, and for comparison for a commercial Cuprophan membrane. Pore radii of the membranes were determined from these results by three methods: (1) L_p method that uses water hydraulic permeability coefficient, (2) σ method that uses reflection coefficients, and (3) P/L_p method that uses water diffusive permeability coefficient and water hydraulic permeability coefficient.     

Impact of thermal ageing on sorption and diffusion properties of PTMSP

Gas and vapour permeability in both freshly cast and aged poly(1-trimethylsilyl-1-propyne) (PTMSP) membranes were investigated in terms of solubility and diffusion coefficients for two probe molecules, a permanent gas (nitrogen) and an organic vapour (dichloromethane). To get reliable data for this study, we set up a fast and reproducible ageing procedure consisting of thermal treatment of the polymer films (100 °C during 24 h under vacuum). As expected, measurements recorded from time-lag experiments and isothermal sorption showed strong variations of the PTMSP transport properties before and after the thermal ageing procedure. Freshly cast membranes exhibited high permeability, whereas after ageing a 40–45% decrease of the permeability was recorded for both probes. The results demonstrated that only the glassy physical microstructure of PTMSP was affected by the ageing procedure, while the chemical structure was unchanged. Based on a dual-mode model for sorption and a Long's model for diffusion, the analysis of the data showed that the solubility and diffusion coefficients of the gas and the vapour were not affected in the same way. For nitrogen, only the diffusion coefficient decreased, whereas for dichloromethane, the thermal treatment mainly influenced the sorption coefficient. The lower permeability due to the combination of sorption and diffusion parameters could be attributed to a change of the PTMSP hole geometry or the hole connections.     

DeepReac+: deep active learning for quantitative modeling of organic chemical reactions

Various computational methods have been developed for quantitative modeling of organic chemical reactions; however, the lack of universality as well as the requirement of large amounts of experimental data limit their broad applications. Here, we present DeepReac+, an efficient and universal computational framework for prediction of chemical reaction outcomes and identification of optimal reaction conditions based on deep active learning. Under this framework, DeepReac is designed as a graph-neural-network-based model, which directly takes 2D molecular structures as inputs and automatically adapts to different prediction tasks. In addition, carefully-designed active learning strategies are incorporated to substantially reduce the number of necessary experiments for model training. We demonstrate the universality and high efficiency of DeepReac+ by achieving the state-of-the-art results with a minimum of labeled data on three diverse chemical reaction datasets in several scenarios. Collectively, DeepReac+ has great potential and utility in the development of AI-aided chemical synthesis. DeepReac+ is freely accessible at https://github.com/bm2-lab/DeepReac.

Based on GNNs and active learning, DeepReac+ is designed as a universal framework for quantitative modeling of chemical reactions. It takes molecular structures as inputs directly and adapts to various prediction tasks with fewer training data.     

Semi-correlations as a tool to build up categorical (active/inactive) model of GABAA receptor modulator activity

The interpretation of mode of action for GABA_A receptor modulator activity is an important task of medicinal chemistry. The computational elucidation of the modulator activity is one of the ways to solve the above task. So-called semi-correlation is a tool for prediction of GABA_A receptor modulator activity. The semi-correlation is based on the Monte Carlo method. This approach is to build up categorical classification models into two classes: (i) active and (ii) inactive. The CORAL software (http://www.insilico.eu/coral) can be used to build up the semi-correlations. The statistical quality of models (for external validation sets) based on semi-correlation has the range of Matthews correlation coefficient (MCC) is 0.72–1.00 for 30 random splits of all available data (n?=?210) into the training and validation sets. In contrast to existing approaches, the predictive CORAL models give prediction using solely data on molecular architecture (represented by simplified molecular input-line entry system?=?SMILES) and available experimental data on endpoints. Suggested models for prediction of GABA_A receptor modulator activity are built up according to the OECD principles. Thus, the approach based on the semi-correlation can be a useful tool for studying of the GABA_A receptor modulators activity.

     

A sequence-based computational model for the prediction of the solvent accessible surface area for α-helix and β-barrel transmembrane residues

Predicting the solvent accessible surface area (ASA) of transmembrane (TM) residues is of great importance for experimental researchers to elucidate diverse physiological processes. TM residues fall into two major structural classes (α-helix membrane protein and β-barrel membrane protein). The reported solvent ASA prediction models were developed for these two types of TM residues respectively. However, this prevents the general use of these methods because one cannot determine which model is suitable for a given TM residue without information of its type. To conquer this limitation, we developed a new computational model that can be used for predicting the ASA of both TM α-helix and β-barrel residues. The model was developed from 78 α-helix membrane protein chains and 24 β-barrel membrane protein. Its prediction ability was evaluated by cross validation method and its prediction result on an independent test set of 20 membrane protein chains. The results show that our model performs well for both types of TM residues and outperforms other prediction model which was developed for the specific type of TM residues. The prediction results also proved that the random forest model incorporating conservation score is an effective sequence-based computational approach for predicting the solvent ASA of TM residues.