Development of Novel Methods for QSAR Modeling by Machine Learning Repeatedly: A Case Study on Drug Distribution to Each Tissue.

Artificial intelligence is expected to help identify excellent candidates in drug discovery. However, we face a lack of data, as it is time-consuming and expensive to acquire raw data perfectly for many compounds. Hence, we tried to develop a novel quantitative structure-activity relationship (QSAR) method to predict a parameter more precisely from an incomplete data set via optimizing data handling by making use of predicted explanatory variables. As a case study we focused on the tissue-to-plasma partition coefficient (Kp), which is an important parameter for understanding drug distribution in tissues and building the physiologically based pharmacokinetic model and is a representative of small and sparse data sets. In this study, we predicted the Kp values of 119 compounds in nine tissues (adipose, brain, gut, heart, kidney, liver, lung, muscle, and skin), although some of these were not available. To fill the missing values in Kp for each tissue, first we predicted those Kp values by the nonmissing data set using a random forest (RF) model with in vitro parameters (log P, fu, Drug Class, and fi) like a classical prediction by a QSAR model. Next, to predict the tissue-specific Kp values in a test data set, we constructed a second RF model with not only in vitro parameters but also the Kp values of other tissues (i.e., other than target tissues) predicted by the first RF model as explanatory variables. Furthermore, we tested all possible combinations of explanatory variables and selected the model with the highest predictability from the test data set as the final model. The evaluation of Kp prediction accuracy based on the root-mean-square error and R2 value revealed that the proposed models outperformed other machine learning methods such as the conventional RF and message-passing neural networks. Significant improvements were observed in the Kp values of adipose tissue, brain, kidney, liver, and skin. These improvements indicated that the Kp information on other tissues can be used to predict the same for a specific tissue. Additionally, we found a novel relationship between each tissue by evaluating all combinations of explanatory variables. In conclusion, we developed a novel RF model to predict Kp values. We hope that this method will be applied to various problems in the field of experimental biology which often contains missing values in the near future.

Prediction of Compound Plasma Concentration-Time Profiles in Mice Using Random Forest.

Pharmacokinetic (PK) parameters such as clearance (CL) and volume of distribution (Vd) have been the subject of previous in silico predictive models. However, having information of the concentration over time profile explicitly can provide additional value like time above MIC or AUC, etc., to understand both the efficacy and safety-related aspects of a compound. In this work, we developed machine learning models for plasma concentration-time profiles after both i.v. and p.o. dosing for a series of 17 in-house projects. For explanatory variables, MACCS Keys chemical descriptors as well as in silico and experimental in vitro PK parameters were used. The predictive accuracy of random forest (RF), message passing neural network, 2-compartment models using estimated CL and Vdss, and an average model (as a control experiment) was investigated using 5-fold cross-validation (5-fold CV) and leave-one-project-out validation (LOPO-V). The predictive accuracy of RF in 5-fold CV for i.v. and p.o. plasma concentration-time profiles was the best among the models studied, with an RMSE for i.v. dosing at 0.08, 1, and 8 h of 0.245, 0.474, and 0.462, respectively, and an RMSE for p.o. dosing at 0.25, 1, and 8 h of 0.500, 0.612, and 0.509, respectively. Furthermore, by investigating the importance of the in vitro PK parameters using the Gini index, we observed that the general prior knowledge in ADME research was reflected well in the respective feature importance of in vitro parameters such as predicted human Vd (hVd) for the initial distribution, mouse intrinsic CL and unbound fraction of mouse plasma for the elimination process, and Caco2 permeability for the absorption process. Also, this model is the first model that can predict twin peaks in the concentration-time profile much better than a baseline compartment model. Because of its combination of sufficient accuracy and speed of prediction, we found the model to be fit-for-purpose for practical lead optimization.

Predicting Total Drug Clearance and Volumes of Distribution Using the Machine Learning-Mediated Multimodal Method through the Imputation of Various Nonclinical Data.

Pharmacokinetic research plays an important role in the development of new drugs. Accurate predictions of human pharmacokinetic parameters are essential for the success of clinical trials. Clearance (CL) and volume of distribution (Vd) are important factors for evaluating pharmacokinetic properties, and many previous studies have attempted to use computational methods to extrapolate these values from nonclinical laboratory animal models to human subjects. However, it is difficult to obtain sufficient, comprehensive experimental data from these animal models, and many studies are missing critical values. This means that studies using nonclinical data as explanatory variables can only apply a small number of compounds to their model training. In this study, we perform missing-value imputation and feature selection on nonclinical data to increase the number of training compounds and nonclinical datasets available for these kinds of studies. We could obtain novel models for total body clearance (CLtot) and steady-state Vd (Vdss) (CLtot: geometric mean fold error [GMFE], 1.92; percentage within 2-fold error, 66.5%; Vdss: GMFE, 1.64; percentage within 2-fold error, 71.1%). These accuracies were comparable to the conventional animal scale-up models. Then, this method differs from animal scale-up methods because it does not require animal experiments, which continue to become more strictly regulated as time passes.

Development of a Novel In Silico Classification Model to Assess Reactive Metabolite Formation in the Cysteine Trapping Assay and Investigation of Important Substructures.

Predicting whether a compound can cause drug-induced liver injury (DILI) is difficult due to the complexity of drug mechanism. The cysteine trapping assay is a method for detecting reactive metabolites that bind to microsomes covalently. However, it is cumbersome to use 35S isotope-labeled cysteine for this assay. Therefore, we constructed an in silico classification model for predicting a positive/negative outcome in the cysteine trapping assay. We collected 475 compounds (436 in-house compounds and 39 publicly available drugs) based on experimental data performed in this study, and the composition of the results showed 248 positives and 227 negatives. Using a Message Passing Neural Network (MPNN) and Random Forest (RF) with extended connectivity fingerprint (ECFP) 4, we built machine learning models to predict the covalent binding risk of compounds. In the time-split dataset, AUC-ROC of MPNN and RF were 0.625 and 0.559 in the hold-out test, restrictively. This result suggests that the MPNN model has a higher predictivity than RF in the time-split dataset. Hence, we conclude that the in silico MPNN classification model for the cysteine trapping assay has a better predictive power. Furthermore, most of the substructures that contributed positively to the cysteine trapping assay were consistent with previous results.

Development of a 2D-QSAR Model for Tissue-to-Plasma Partition Coefficient Value with High Accuracy Using Machine Learning Method, Minimum Required Experimental Values, and Physicochemical Descriptors.

BACKGROUND: The demand for physiologically based pharmacokinetic (PBPK) model is increasing currently. New drug application (NDA) of many compounds is submitted with PBPK models for efficient drug development. Tissue-to-plasma partition coefficient (Kp) is a key parameter for the PBPK model to describe differential equations. However, it is difficult to obtain the Kp value experimentally because the measurement of drug concentration in the tissue is much harder than that in plasma. OBJECTIVE: Instead of experiments, many researchers have sought in silico methods. Today, most of the models for Kp prediction are using in vitro and in vivo parameters as explanatory variables. We thought of physicochemical descriptors that could improve the predictability. Therefore, we aimed to develop the two-dimensional quantitative structure-activity relationship (2D-QSAR) model for Kp using physicochemical descriptors instead of in vivo experimental data as explanatory variables. METHODS: We compared our model with the conventional models using 20-fold cross-validation according to the published method (Yun et al. J Pharmacokinet Pharmacodyn 41:1-14, 2014). We used random forest algorithm, which is known to be one of the best predictors for the 2D-QSAR model. Finally, we combined minimum in vitro experimental values and physiochemical descriptors. Thus, the prediction method for Kp value using a few in vitro parameters and physicochemical descriptors was developed; this is a multimodal model. RESULTS: Its accuracy was found to be superior to that of the conventional models. Results of this research suggest that multimodality is useful for the 2D-QSAR model [RMSE and % of two-fold error: 0.66 and 42.2% (Berezohkovsky), 0.52 and 52.2% (Rodgers), 0.65 and 34.6% (Schmitt), 0.44 and 61.1% (published model), 0.41 and 62.1% (traditional model), 0.39 and 64.5% (multimodal model)]. CONCLUSION: We could develop a 2D-QSAR model for Kp value with the highest accuracy using a few in vitro experimental data and physicochemical descriptors.

On the difficulty of validating molecular generative models realistically: a case study on public and proprietary data.

While a multitude of deep generative models have recently emerged there exists no best practice for their practically relevant validation. On the one hand, novel de novo-generated molecules cannot be refuted by retrospective validation (so that this type of validation is biased); but on the other hand prospective validation is expensive and then often biased by the human selection process. In this case study, we frame retrospective validation as the ability to mimic human drug design, by answering the following question: Can a generative model trained on early-stage project compounds generate middle/late-stage compounds de novo? To this end, we used experimental data that contains the elapsed time of a synthetic expansion following hit identification from five public (where the time series was pre-processed to better reflect realistic synthetic expansions) and six in-house project datasets, and used REINVENT as a widely adopted RNN-based generative model. After splitting the dataset and training REINVENT on early-stage compounds, we found that rediscovery of middle/late-stage compounds was much higher in public projects (at 1.60%, 0.64%, and 0.21% of the top 100, 500, and 5000 scored generated compounds) than in in-house projects (where the values were 0.00%, 0.03%, and 0.04%, respectively). Similarly, average single nearest neighbour similarity between early- and middle/late-stage compounds in public projects was higher between active compounds than inactive compounds; however, for in-house projects the converse was true, which makes rediscovery (if so desired) more difficult. We hence show that the generative model recovers very few middle/late-stage compounds from real-world drug discovery projects, highlighting the fundamental difference between purely algorithmic design and drug discovery as a real-world process. Evaluating de novo compound design approaches appears, based on the current study, difficult or even impossible to do retrospectively.Scientific Contribution This contribution hence illustrates aspects of evaluating the performance of generative models in a real-world setting which have not been extensively described previously and which hopefully contribute to their further future development.

Prediction of Total Drug Clearance in Humans Using Animal Data: Proposal of a Multimodal Learning Method Based on Deep Learning.

Research into pharmacokinetics plays an important role in the development process of new drugs. Accurately predicting human pharmacokinetic parameters from preclinical data can increase the success rate of clinical trials. Since clearance (CL) which indicates the capacity of the entire body to process a drug is one of the most important parameters, many methods have been developed. However, there are still rooms to be improved for practical use in drug discovery research; "improving CL prediction accuracy" and "understanding the chemical structure of compounds in terms of pharmacokinetics". To improve those, this research proposes a multimodal learning method based on deep learning that takes not only the chemical structure of a drug but also rat CL as inputs. Good results were obtained compared with the conventional animal scale-up method; the geometric mean fold error was 2.68 and the proportion of compounds with prediction errors of 2-fold or less was 48.5%. Furthermore, it was found to be possible to infer the partial structure useful for CL prediction by a structure contributing factor inference method. The validity of these results of structural interpretation of metabolic stability was confirmed by chemists.

Calcium ions are involved in the delay of plant cell cycle progression by abiotic stresses.

Higher plants respond to environmental stresses by a sequence of reactions which include the reduction of growth by affecting cell division. It has been shown that calcium ions plays a role as a second messenger in mediating various defence responses under environmental stresses. In this study, the role of calcium ions on cell cycle progression under abiotic stresses has been examined in tobacco BY-2 suspension culture cells. Using synchronized BY-2 cells expressing the endogenous calcium sensor aequorin as experimental system, we could show that oxidative and hypoosmotic stress both induce an increase of intracellular calcium and cause a delay of the cell cycle. The inhibitory effect of these abiotic stress stimuli on cell cycle progression could be mimicked by increasing the intracellular calcium concentration via application of an external electrical field. Likewise, depletion of calcium ions in the culture medium suppressed the effect of the stimuli tested. These results demonstrate that calcium signalling is involved in the regulation of cell cycle progression in response to abiotic stress.

Three-dimensional quantitative structure-activity relationship analysis for human pregnane X receptor for the prediction of CYP3A4 induction in human hepatocytes: structure-based comparative molecular field analysis.

The pregnane X receptor [PXR (NR1I2)] induces the expression of xenobiotic metabolic genes and transporter genes. In this study, we aimed to establish a computational method for quantifying the enzyme-inducing potencies of different compounds via their ability to activate PXR, for the application in drug discovery and development. To achieve this purpose, we developed a three-dimensional quantitative structure-activity relationship (3D-QSAR) model using comparative molecular field analysis (CoMFA) for predicting enzyme-inducing potencies, based on computer-ligand docking to multiple PXR protein structures sampled from the trajectory of a molecular dynamics simulation. Molecular mechanics-generalized born/surface area scores representing the ligand-protein-binding free energies were calculated for each ligand. As a result, the predicted enzyme-inducing potencies for compounds generated by the CoMFA model were in good agreement with the experimental values. Finally, we concluded that this 3D-QSAR model has the potential to predict the enzyme-inducing potencies of novel compounds with high precision and therefore has valuable applications in the early stages of the drug discovery process.

In silico study on the inhibitory interaction of drugs with wild-type CYP2D6.1 and the natural variant CYP2D6.17.

The natural variant of the cytochrome P450 enzyme CYP2D6.1, CYP2D6.17, is most common in African populations, has three amino acid substitutions (T107I, R296C, and S486T) compared to the wild-type, and is known to have a different ligand preference from CYP2D6.1. It is becoming increasingly important to understand differences in the metabolism of medicines in different ethnic groups in order to assess the relevance of clinical data from different countries. This study investigated differences in the inhibition profiles of drugs for CYP2D6 with respect to gene polymorphisms. Firstly, we used computer docking with six drugs to several CYP2D6.1 structures, sampled from the trajectory of MD simulations, and calculated MM-GB/SA scores representing binding free energies. We then used regression analysis to predict the potency with which drugs inhibited CYP2D6.1 based on MM-GB/SA scores. The pKi-values obtained were in good agreement with experimental values measured for the six drugs (r(2) = 0.81). We carried out the same analysis for CYP2D6.17 and the pKi-values calculated were also in good agreement with experimental values (r(2) = 0.92). Finally, we were able to successfully explain the different abilities of CYP2D6.1 and CYP2D6.17 to metabolize drugs in different ethnic groups with reference to their 3D-structures.

Three-dimensional quantitative structure-activity relationship analysis of inhibitors of human and rat cytochrome P4503A enzymes.

Cytochrome P450 3A4 (CYP3A4) is a member of the CYP family and is an important enzyme in drug metabolism. A compound that inhibits CYP3A4 activity could also affect the pharmacokinetics of other substrates, resulting in drug-drug interactions (DDIs) that could cause side effects. Pharmacokinetic data from drug-development studies in rats often determine the dosage used in human clinical trials. It is therefore useful to understand differences in metabolism in different species at an early stage in drug development. Human and rat CYP3A enzymes show different inhibition profiles with different drugs, although the mechanisms involved are not yet clear. Here we built three-dimensional quantitative structure-activity relationship (3D-QSAR) models using structure-based comparative molecular field analysis (CoMFA), to predict the direct inhibitory activity of ligands for human CYP3A4 and rat CYP3A1, based on computer-ligand docking. The alignment of the ligand docking poses suggested that key amino acid-ligand interactions (e.g., Thr309 in CYP3A4 and Pro310 in CYP3A1) characterized the different potencies with which the ligands inhibited CYP3A4 and CYP3A1. The 3D-QSAR models for human and rat CYP3A family inhibitors predicted the potency of inhibitors and could be useful for assessing DDIs at an early stage in drug discovery.