In silico modeling-based new alternative methods to predict drug and herb-induced liver injury: A review.

New approach methods (NAMs) have been developed to predict a wide range of toxicities through innovative technologies. Liver injury is one of the most extensively studied endpoints due to its severity and frequency, occurring among populations that consume drugs or dietary supplements. In this review, we focus on recent developments of in silico modeling for liver injury prediction using deep learning and in vitro data based on adverse outcome pathways (AOPs). Despite these models being mainly developed using datasets generated from drug-like molecules, they were also applied to the prediction of hepatotoxicity caused by herbal products. As deep learning has achieved great success in many different fields, advanced machine learning algorithms have been actively applied to improve the accuracy of in silico models. Additionally, the development of liver AOPs, combined with big data in toxicology, has been valuable in developing in silico models with enhanced predictive performance and interpretability. Specifically, one approach involves developing structure-based models for predicting molecular initiating events of liver AOPs, while others use in vitro data with structure information as model inputs for making predictions. Even though liver injury remains a difficult endpoint to predict, advancements in machine learning algorithms and the expansion of in vitro databases with relevant biological knowledge have made a huge impact on improving in silico modeling for drug-induced liver injury prediction.

Semi-automated approach for generation of biological networks on drug-induced cholestasis, steatosis, hepatitis, and cirrhosis.

Drug-induced liver injury (DILI) is one of the leading reasons for discontinuation of a new drug development project. Diverse machine learning or deep learning models have been developed to predict DILI. However, these models have not provided an adequate understanding of the mechanisms leading to DILI. The development of safer drugs requires novel computational approaches that enable the prompt understanding of the mechanism of DILI. In this study, the mechanisms leading to the development of cholestasis, steatosis, hepatitis, and cirrhosis were explored using a semi-automated approach for data gathering and associations. Diverse data from ToxCast, Comparative Toxicogenomic Database (CTD), Reactome, and Open TG-GATEs on reference molecules leading to the development of the respective diseases were extracted. The data were used to create biological networks of the four diseases. As expected, the four networks had several common pathways, and a joint DILI network was assembled. Such biological networks could be used in drug discovery to identify possible molecules of concern as they provide a better understanding of the disease-specific key events. The events can be target-tested to provide indications for potential DILI effects. Supplementary Information: The online version contains supplementary material available at 10.1007/s43188-022-00124-6.

ToxSTAR: drug-induced liver injury prediction tool for the web environment.

SUMMARY: Drug-induced liver injury (DILI) is a challenging endpoint in predictive toxicology because of the complex reactive metabolites that cause liver damage and the wide range of mechanisms involved in the development of the disease. ToxSTAR provides structural similarity-based DILI analysis and in-house DILI prediction models that predict four DILI subtypes (cholestasis, cirrhosis, hepatitis and steatosis) based on drug and drug metabolite molecules. AVAILABILITY AND IMPLEMENTATION: ToxSTAR is freely available at https://toxstar.kitox.re.kr/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Development of an Infinite Dilution Activity Coefficient Prediction Model for Organic Solutes in Ionic Liquids with Modified Partial Equalization Orbital Electronegativity Method Derived Descriptors.

The objective of this study was to develop a robust prediction model for the infinite dilution activity coefficients (γ ∞) of organic molecules in diverse ionic liquid (IL) solvents. Electrostatic, hydrogen bond, polarizability, molecular structure, and temperature terms were used in model development. A feed-forward model based on artificial neural networks was developed with 34,754 experimental activity coefficients, a combination of 195 IL solvents (88 cations and 38 anions), and 147 organic solutes at a temperature range of 298 to 408 K. The root mean squared error (RMSE) of the training set and test set was 0.219 and 0.235, respectively. The R 2 of the training set and the test set was 0.984 and 0.981, respectively. The applicability domain was determined through a Williams plot, which implied that water and halogenated compounds were outside of the applicability domain. The robustness test shows that the developed model is robust. The web server supports using the developed prediction model and is freely available at https://preadmet.bmdrc.kr/activitycoefficient_mainpage/prediction/.

Nano-QTTR development for interspecies aquatic toxicity of silver nanoparticles between daphnia and fish.

Limited studies of quantitative toxicity-toxicity relationship (QTTR) modeling have been conducted to predict interspecies toxicity of engineered nanomaterials (ENMs) between aquatic test species. A meta-analysis of 66 publications providing acute toxicity data of silver nanoparticles (AgNPs) to daphnia and fish was performed, and the toxicity data, physicochemical properties, and experimental conditions were collected and curated. Based on Euclidean distance (ED) grouping, a meaningful correlation of logarithmic lethal concentrations between daphnia and fish was derived for bare (R2bare = 0.47) and coated AgNPs (R2coated = 0.48) when a distance of 10 was applied. The correlation of coated AgNPs was improved (R2coated = 0.55) by the inclusion of descriptors of the coating materials. The correlations were further improved by R2bare = 0.57 and R2coated = 0.81 after additionally considering particle size only, and by R2bare = 0.59 and R2coated = 0.92 after considering particle size and zeta potential simultaneously. The developed ED-based nano-QTTR model demonstrated that inclusion of the coating material descriptors and physicochemical properties improved the goodness-of-fit to predict interspecies aquatic toxicity of AgNPs between daphnia and fish. This study provides insight for future in silico research on QTTR model development in ENM toxicology.

Development of blood brain barrier permeation prediction models for organic and inorganic biocidal active substances.

Biocidal products are broadly used in homes and industries. However, the safety of biocidal active substances (BASs) is not yet fully understood. In particular, the neurotoxic action of BASs needs to be studied as diverse epidemiological studies have reported associations between exposure to BASs and neural diseases. In this study, we developed in silico models to predict the blood-brain barrier (BBB) permeation of organic and inorganic BASs. Due to a lack of BBB data for BASs, the chemical space of BASs and BBB dataset were compared in order to select BBB data that were structurally similar to BASs. In silico models to predict log-scaled BBB penetration were developed using support vector regression for organic BASs and multiple linear regression for inorganic BASs. The model for organic BASs was developed with 231 compounds (training set: 153 and test set: 78) and achieved good prediction accuracy on an external test set (R2 = 0.64), and the model outperformed the model for pharmaceuticals. The model for inorganic BASs was developed with 11 compounds (R2 = 0.51). Applicability domain (AD) analysis of the models clarified molecular structures reliably predicted by the models. Therefore, the models developed in this study can be used for predicting BBB permeable BASs in human. These models were developed according to the Quantitative Structure-Activity Relationship validation principles proposed by the Organization for Economic Cooperation and Development.

Use of size-dependent electron configuration fingerprint to develop general prediction models for nanomaterials.

Due to the lack of nano descriptors that can appropriately represent the wide chemical space of engineered nanomaterials (ENMs), applicability domain of nano-quantitative structure-activity relationship models are limited to certain types of ENMs, such as metal oxides, metals, carbon-based nanomaterials, or quantum dots. In this study, a size-dependent electron configuration fingerprint (SDEC FP) was introduced to estimate the quantity of electrons based on the core, doping, and coating materials of ENMs in different sizes. SDEC FP was used in prediction model development and nanostructure similarity analysis on datasets including metal and carbon-based nanomaterials with and without surface modifications. Cytotoxicity and zeta potential prediction models developed with SDEC FP achieved good prediction accuracies on test set. Nanostructure similarity analysis was performed through principal component analysis which showed that structural similarity between ENMs measured by SDEC FP was highly correlated with their properties.

Topological Distance-Based Electron Interaction Tensor to Apply a Convolutional Neural Network on Drug-like Compounds.

Deep learning (DL) models in quantitative structure-activity relationship fed the molecular structure directly to the network without using human-designed descriptors by representing molecule as a graph or string (e.g., SMILES code). However, these two representations were oversimplification of real molecules to reflect chemical properties of molecular structures. Given that the choice of molecular representation determines the architecture of the DL model to apply, a novel way of molecular representation can open a way to apply diverse DL networks developed and used in other fields. A topological distance-based electron interaction (TDEi) tensor has been developed in this study inspired by the quantum mechanical model of the molecule, which defines a molecule with electrons and protons. In the TDEi tensor, the atomic orbital (AO) of each atom is represented by an electron configuration (EC) vector, which is a bit string based on the presence and absence of electrons in each AO according to spin indicated by positive and negative signs. Interactions between EC vectors were calculated based on the topological distance between atoms in a molecule. As a molecular structure was translated into 3D array, CNN models (modified VGGNet) were applied using a TDEi tensor to predict four physicochemical properties of drug-like compound datasets: MP (275,131), Lipop (4193), Esol (1127), and Freesolv (639). Models achieved good prediction accuracy. PCA showed that a stronger correlation was observed between the extracted features and the target endpoint as features were extracted from the deeper layer.

PreMetabo: An in silico phase I and II drug metabolism prediction platform.

This study aimed to develop a drug metabolism prediction platform using knowledge-based prediction models. Site of Metabolism (SOM) prediction models for four cytochrome P450 (CYP) subtypes were developed along with uridine 5'-diphosphoglucuronosyltransferase (UGT) and sulfotransferase (SULT) substrate classification models. The SOM substrate for a certain CYP was determined using the sum of the activation energy required for the reaction at the reaction site of the substrate and the binding energy of the substrate to the CYP enzyme. Activation energy was calculated using the EaMEAD model and binding energy was calculated by docking simulation. Phase II prediction models were developed to predict whether a molecule is the substrate of a certain phase II conjugate protein, i.e., UGT or SULT. Using SOM prediction models, the predictability of the major metabolite in the top-3 was obtained as 72.5-84.5% for four CYPs, respectively. For internal validation, the accuracy of the UGT and SULT substrate classification model was obtained as 93.94% and 80.68%, respectively. Additionally, for external validation, the accuracy of the UGT substrate classification model was obtained as 81% in the case of 11 FDA-approved drugs. PreMetabo is implemented in a web environment and is available at https://premetabo.bmdrc.kr/.

Development of Prediction Models for Drug-Induced Cholestasis, Cirrhosis, Hepatitis, and Steatosis Based on Drug and Drug Metabolite Structures.

Drug-induced liver injury (DILI) is one of the major reasons for termination of drug development. Due to the importance of predicting DILI in early phases of drug development, diverse in silico models have been developed to filter out DILI-causing candidates before clinical study. However, no computational models have achieved sufficient prediction power for screening DILI in early phases because 1) drugs often cause liver injury through reactive metabolites, 2) different clinical outcomes of DILI have different mechanisms, and 3) the DILI label on drugs is not clearly defined. In this study, we developed binary classification models to predict drug-induced cholestasis, cirrhosis, hepatitis, and steatosis based on the structure of drugs and their metabolites. DILI-positive data was obtained from post-market reports of drugs and DILI-negative data from DILIrank, a database curated by the Food and Drug Administration (FDA). Support vector machine (SVM) and random forest (RF) were used in developing models with nine fingerprints and one 2D molecular descriptor calculated from drug (152 DILI-positives and 102 DILI-negatives) and drug metabolite (192 DILI-positives and 126 DILI-negatives) structures. Models were developed according to Organisation for Economic Co-operation and Development (OECD) guidelines for quantitative structure-activity relationship (QSAR) validation. Internal and external validation was performed with a randomization test in order to thoroughly examine model predictability and avoid random correlation between structural features and adverse outcomes. The applicability domain was defined with a leverage method for reliable prediction of new chemicals. The best models for each liver disease were selected based on external validation results from drugs (cholestasis: 70%, cirrhosis: 90%, hepatitis: 83%, and steatosis: 85%) and drug metabolites (cholestasis: 86%, cirrhosis: 88%, hepatitis: 86%, and steatosis: 83%) with applicability domain analysis. Compiled data sets were further exploited to derive privileged substructures that were more frequent in DILI-positive sets compared to DILI-negative sets and in drug metabolite structures compared to drug structures with a Morgan fingerprint level 2.

Electron configuration-based neural network model to predict physicochemical properties of inorganic compounds.

Registration, evaluation, and authorization of chemicals (REACH), the regulation of chemicals in use, imposes the characterization and report of the physicochemical properties of compounds. To cope with the financial burden of the experiments, the use of computational models is permitted for prediction of properties. Although a number of physicochemical property prediction models have been developed, their applicability domain is limited to organic molecules since most available data are concerned with organic molecules, and most of the molecular descriptors are restricted to organic molecule calculations. Prediction models developed for inorganic compounds were intended to predict endpoints relevant to novel material design. Therefore, no models were available for predicting endpoints of inorganic compounds that are significant to regulatory perspectives. In this study, boiling point, water solubility, melting point, and pyrolysis point prediction models were developed for inorganic compounds based on their composition. The electron configuration of each element in the molecule was used as a descriptor in this study. The dataset covered a wide range of endpoints and diverse elements in their structure. The performance of the models was measured using R 2, mean absolute error, and Spearman's correlation coefficient, and indicated good prediction accuracy of continuous endpoints and prioritization of inorganic compounds.

Development of Natural Compound Molecular Fingerprint (NC-MFP) with the Dictionary of Natural Products (DNP) for natural product-based drug development.

Computer-aided research on the relationship between molecular structures of natural compounds (NC) and their biological activities have been carried out extensively because the molecular structures of new drug candidates are usually analogous to or derived from the molecular structures of NC. In order to express the relationship physically realistically using a computer, it is essential to have a molecular descriptor set that can adequately represent the characteristics of the molecular structures belonging to the NC's chemical space. Although several topological descriptors have been developed to describe the physical, chemical, and biological properties of organic molecules, especially synthetic compounds, and have been widely used for drug discovery researches, these descriptors have limitations in expressing NC-specific molecular structures. To overcome this, we developed a novel molecular fingerprint, called Natural Compound Molecular Fingerprints (NC-MFP), for explaining NC structures related to biological activities and for applying the same for the natural product (NP)-based drug development. NC-MFP was developed to reflect the structural characteristics of NCs and the commonly used NP classification system. NC-MFP is a scaffold-based molecular fingerprint method comprising scaffolds, scaffold-fragment connection points (SFCP), and fragments. The scaffolds of the NC-MFP have a hierarchical structure. In this study, we introduce 16 structural classes of NPs in the Dictionary of Natural Product database (DNP), and the hierarchical scaffolds of each class were calculated using the Bemis and Murko (BM) method. The scaffold library in NC-MFP comprises 676 scaffolds. To compare how well the NC-MFP represents the structural features of NCs compared to the molecular fingerprints that have been widely used for organic molecular representation, two kinds of binary classification tasks were performed. Task I is a binary classification of the NCs in commercially available library DB into a NC or synthetic compound. Task II is classifying whether NCs with inhibitory activity in seven biological target proteins are active or inactive. Two tasks were developed with some molecular fingerprints, including NC-MFP, using the 1-nearest neighbor (1-NN) method. The performance of task I showed that NC-MFP is a practical molecular fingerprint to classify NC structures from the data set compared with other molecular fingerprints. Performance of task II with NC-MFP outperformed compared with other molecular fingerprints, suggesting that the NC-MFP is useful to explain NC structures related to biological activities. In conclusion, NC-MFP is a robust molecular fingerprint in classifying NC structures and explaining the biological activities of NC structures. Therefore, we suggest NC-MFP as a potent molecular descriptor of the virtual screening of NC for natural product-based drug development.