Application of in silico approaches to predicting drug--drug interactions

In silico ADME/tox comes of age: twenty years later

In the early 2000s pharmaceutical drug discovery was beginning to use computational approaches for absorption, distribution, metabolism, excretion and toxicity (ADME/Tox, also known as ADMET) prediction. This emphasis on prediction was an effort to reduce the risk of later stage failures from ADME/Tox.Much has been written in the intervening twenty plus years and significant expenditure has occurred in companies developing these in silico capabilities which can be gleaned from publications. It is therefore an appropriate time to briefly reflect on what was proposed then and what the reality is today.20 years ago, we tended to optimise bioactivity and perhaps one ADME/Tox property at a time. Previously pharmaceutical companies needed a whole infrastructure for models - in silico and in vitro experts, IT, champions on a project team, educators and management support. Now we are in the age of generative de novo design where bioactivity and many ADME/Tox properties can be optimised and large language model technologies are available.There are also some challenges such as the focus on very large molecules which may be outside of current ADME/Tox models.We provide an opportunity to look forward with the increasing public data for ADME/Tox as well as expanded types of algorithms available.

Machine Learning Models Identify New Inhibitors for Human OATP1B1

The uptake transporter OATP1B1 (SLC01B1) is largely localized to the sinusoidal membrane of hepatocytes and is a known victim of unwanted drug-drug interactions. Computational models are useful for identifying potential substrates and/or inhibitors of clinically relevant transporters. Our goal was to generate OATP1B1 in vitro inhibition data for [3H] estrone-3-sulfate (E3S) transport in CHO cells and use it to build machine learning models to facilitate a comparison of seven different classification models (Deep learning, Adaboosted decision trees, Bernoulli naïve bayes, k-nearest neighbors (knn), random forest, support vector classifier (SVC), logistic regression (lreg), and XGBoost (xgb)] using ECFP6 fingerprints to perform 5-fold, nested cross validation. In addition, we compared models using 3D pharmacophores, simple chemical descriptors alone or plus ECFP6, as well as ECFP4 and ECFP8 fingerprints. Several machine learning algorithms (SVC, lreg, xgb, and knn) had excellent nested cross validation statistics, particularly for accuracy, AUC, and specificity. An external test set containing 207 unique compounds not in the training set demonstrated that at every threshold SVC outperformed the other algorithms based on a rank normalized score. A prospective validation test set was chosen using prediction scores from the SVC models with ECFP fingerprints and were tested in vitro with 15 of 19 compounds (84% accuracy) predicted as active (≥20% inhibition) showed inhibition. Of these compounds, six (abamectin, asiaticoside, berbamine, doramectin, mobocertinib, and umbralisib) appear to be novel inhibitors of OATP1B1 not previously reported. These validated machine learning models can now be used to make predictions for drug-drug interactions for human OATP1B1 alongside other machine learning models for important drug transporters in our MegaTrans software.

Predicting Drug-Drug Interactions Based on Integrated Similarity and Semi-Supervised Learning

A drug-drug interaction (DDI) is defined as an association between two drugs where the pharmacological effects of a drug are influenced by another drug. Positive DDIs can usually improve the therapeutic effects of patients, but negative DDIs cause the major cause of adverse drug reactions and even result in the drug withdrawal from the market and the patient death. Therefore, identifying DDIs has become a key component of the drug development and disease treatment. In this study, we propose a novel method to predict DDIs based on the integrated similarity and semi-supervised learning (DDI-IS-SL). DDI-IS-SL integrates the drug chemical, biological and phenotype data to calculate the feature similarity of drugs with the cosine similarity method. The Gaussian Interaction Profile kernel similarity of drugs is also calculated based on known DDIs. A semi-supervised learning method (the Regularized Least Squares classifier) is used to calculate the interaction possibility scores of drug-drug pairs. In terms of the 5-fold cross validation, 10-fold cross validation and de novo drug validation, DDI-IS-SL can achieve the better prediction performance than other comparative methods. In addition, the average computation time of DDI-IS-SL is shorter than that of other comparative methods. Finally, case studies further demonstrate the performance of DDI-IS-SL in practical applications.

Artificial intelligence, machine learning, and drug repurposing in cancer

Introduction: Drug repurposing provides a cost-effective strategy to re-use approved drugs for new medical indications. Several machine learning (ML) and artificial intelligence (AI) approaches have been developed for systematic identification of drug repurposing leads based on big data resources, hence further accelerating and de-risking the drug development process by computational means.Areas covered: The authors focus on supervised ML and AI methods that make use of publicly available databases and information resources. While most of the example applications are in the field of anticancer drug therapies, the methods and resources reviewed are widely applicable also to other indications including COVID-19 treatment. A particular emphasis is placed on the use of comprehensive target activity profiles that enable a systematic repurposing process by extending the target profile of drugs to include potent off-targets with therapeutic potential for a new indication.Expert opinion: The scarcity of clinical patient data and the current focus on genetic aberrations as primary drug targets may limit the performance of anticancer drug repurposing approaches that rely solely on genomics-based information. Functional testing of cancer patient cells exposed to a large number of targeted therapies and their combinations provides an additional source of repurposing information for tissue-aware AI approaches.

Comparison of Deep Learning With Multiple Machine Learning Methods and Metrics Using Diverse Drug Discovery Data Sets

Machine learning methods have been applied to many data sets in pharmaceutical research for several decades. The relative ease and availability of fingerprint type molecular descriptors paired with Bayesian methods resulted in the widespread use of this approach for a diverse array of end points relevant to drug discovery. Deep learning is the latest machine learning algorithm attracting attention for many of pharmaceutical applications from docking to virtual screening. Deep learning is based on an artificial neural network with multiple hidden layers and has found considerable traction for many artificial intelligence applications. We have previously suggested the need for a comparison of different machine learning methods with deep learning across an array of varying data sets that is applicable to pharmaceutical research. End points relevant to pharmaceutical research include absorption, distribution, metabolism, excretion, and toxicity (ADME/Tox) properties, as well as activity against pathogens and drug discovery data sets. In this study, we have used data sets for solubility, probe-likeness, hERG, KCNQ1, bubonic plague, Chagas, tuberculosis, and malaria to compare different machine learning methods using FCFP6 fingerprints. These data sets represent whole cell screens, individual proteins, physicochemical properties as well as a data set with a complex end point. Our aim was to assess whether deep learning offered any improvement in testing when assessed using an array of metrics including AUC, F1 score, Cohen's kappa, Matthews correlation coefficient and others. Based on ranked normalized scores for the metrics or data sets Deep Neural Networks (DNN) ranked higher than SVM, which in turn was ranked higher than all the other machine learning methods. Visualizing these properties for training and test sets using radar type plots indicates when models are inferior or perhaps over trained. These results also suggest the need for assessing deep learning further using multiple metrics with much larger scale comparisons, prospective testing as well as assessment of different fingerprints and DNN architectures beyond those used.

Computational methods in drug discovery

The process for drug discovery and development is challenging, time consuming and expensive. Computer-aided drug discovery (CADD) tools can act as a virtual shortcut, assisting in the expedition of this long process and potentially reducing the cost of research and development. Today CADD has become an effective and indispensable tool in therapeutic development. The human genome project has made available a substantial amount of sequence data that can be used in various drug discovery projects. Additionally, increasing knowledge of biological structures, as well as increasing computer power have made it possible to use computational methods effectively in various phases of the drug discovery and development pipeline. The importance of in silico tools is greater than ever before and has advanced pharmaceutical research. Here we present an overview of computational methods used in different facets of drug discovery and highlight some of the recent successes. In this review, both structure-based and ligand-based drug discovery methods are discussed. Advances in virtual high-throughput screening, protein structure prediction methods, protein-ligand docking, pharmacophore modeling and QSAR techniques are reviewed.

In silico methods for predicting drug-drug interactions with cytochrome P-450s, transporters and beyond

Drug-drug interactions (DDIs) are associated with severe adverse effects that may lead to the patient requiring alternative therapeutics and could ultimately lead to drug withdrawal from the market if they are severe. To prevent the occurrence of DDI in the clinic, experimental systems to evaluate drug interaction have been integrated into the various stages of the drug discovery and development process. A large body of knowledge about DDI has also accumulated through these studies and pharmacovigillence systems. Much of this work to date has focused on the drug metabolizing enzymes such as cytochrome P-450s as well as drug transporters, ion channels and occasionally other proteins. This combined knowledge provides a foundation for a hypothesis-driven in silico approach, using either cheminformatics or physiologically based pharmacokinetics (PK) modeling methods to assess DDI potential. Here we review recent advances in these approaches with emphasis on hypothesis-driven mechanistic models for important protein targets involved in PK-based DDI. Recent efforts with other informatics approaches to detect DDI are highlighted. Besides DDI, we also briefly introduce drug interactions with other substances, such as Traditional Chinese Medicines to illustrate how in silico modeling can be useful in this domain. We also summarize valuable data sources and web-based tools that are available for DDI prediction. We finally explore the challenges we see faced by in silico approaches for predicting DDI and propose future directions to make these computational models more reliable, accurate, and publically accessible.

Open Source Bayesian Models. 1. Application to ADME/Tox and Drug Discovery Datasets

On the order of hundreds of absorption, distribution, metabolism, excretion, and toxicity (ADME/Tox) models have been described in the literature in the past decade which are more often than not inaccessible to anyone but their authors. Public accessibility is also an issue with computational models for bioactivity, and the ability to share such models still remains a major challenge limiting drug discovery. We describe the creation of a reference implementation of a Bayesian model-building software module, which we have released as an open source component that is now included in the Chemistry Development Kit (CDK) project, as well as implemented in the CDD Vault and in several mobile apps. We use this implementation to build an array of Bayesian models for ADME/Tox, in vitro and in vivo bioactivity, and other physicochemical properties. We show that these models possess cross-validation receiver operator curve values comparable to those generated previously in prior publications using alternative tools. We have now described how the implementation of Bayesian models with FCFP6 descriptors generated in the CDD Vault enables the rapid production of robust machine learning models from public data or the user’s own datasets. The current study sets the stage for generating models in proprietary software (such as CDD) and exporting these models in a format that could be run in open source software using CDK components. This work also demonstrates that we can enable biocomputation across distributed private or public datasets to enhance drug discovery.

Novel pre-clinical methodologies for pharmacokinetic drug-drug interaction studies: spotlight on "humanized" animal models

Poly-therapy is common due to co-occurrence of several ailments in patients, leading to the elevated possibility of drug-drug interactions (DDI). Pharmacokinetic DDI often accounts for severe adverse drug reactions in patients resulting in withdrawal of drug from the market. Hence, the prediction of DDI is necessary at pre-clinical stage of drug development. Several human tissue and cell line-based in vitro systems are routinely used for screening metabolic and transporter pathways of investigational drugs and for predicting their clinical DDI potentials. However, ample constraints are associated with the in vitro systems and sometimes in vitro-in vivo extrapolation (IVIVE) fail to assess the risk of DDI in clinic. In vitro-in vivo correlation model in animals combined with human in vitro studies may be helpful in better prediction of clinical outcome. Native animal models vary remarkably from humans in drug metabolizing enzymes and transporters, hence, the interpretation of results from animal DDI studies is difficult. With the advent of modern molecular biology and engineering tools, novel pre-clinical animal models, namely, knockout rat/mouse, transgenic rat/mouse with humanized drug metabolizing enzymes and/or transporters and chimeric rat/mouse with humanized liver are developed. These models nearly simulate human-like drug metabolism and help to validate the in vivo relevance of the in vitro human DDI data. This review briefly discusses the application of such novel pre-clinical models for screening various type of DDI along with their advantages and limitations.

The Therapeutic Potential of hERG1 K+ Channels for Treating Cancer and Cardiac Arrhythmias

hERG potassium channels present pharmacologists and medicinal chemists with a dilemma. On the one hand hERG is a major reason for drugs being withdrawn from the market because of drug induced long QT syndrome and the associated risk of inducing sudden cardiac death, and yet hERG blockers are still widely used in the clinic to treat cardiac arrhythmias. Moreover, in the last decade overwhelming evidence has been provided that hERG channels are aberrantly expressed in cancer cells and that they contribute to tumour cell proliferation, resistance to apoptosis, and neoangiogenesis. Here we provide an overview of the properties of hERG channels and their role in excitable cells of the heart and nervous system as well as in cancer. We consider the therapeutic potential of hERG, not only with regard to the negative impact due to drug induced long QT syndrome, but also its future potential as a treatment in the fight against cancer.

Template-based intervention in Boolean network models of biological systems

MOTIVATION

A grand challenge in the modeling of biological systems is the identification of key variables which can act as targets for intervention. Boolean networks are among the simplest of models, yet they have been shown to adequately model many of the complex dynamics of biological systems. In our recent work, we utilized a logic minimization approach to identify quality single variable targets for intervention from the state space of a Boolean network. However, as the number of variables in a network increases, the more likely it is that a successful intervention strategy will require multiple variables. Thus, for larger networks, such an approach is required in order to identify more complex intervention strategies while working within the limited view of the network's state space. Specifically, we address three primary challenges for the large network arena: the first challenge is how to consider many subsets of variables, the second is to design clear methods and measures to identify the best targets for intervention in a systematic way, and the third is to work with an intractable state space through sampling.

RESULTS

We introduce a multiple variable intervention target called a template and show through simulation studies of random networks that these templates are able to identify top intervention targets in increasingly large Boolean networks. We first show that, when other methods show drastic loss in performance, template methods show no significant performance loss between fully explored and partially sampled Boolean state spaces. We also show that, when other methods show a complete inability to produce viable intervention targets in sampled Boolean state spaces, template methods maintain significantly consistent success rates even as state space sizes increase exponentially with larger networks. Finally, we show the utility of the template approach on a real-world Boolean network modeling T-LGL leukemia.

CONCLUSIONS

Overall, these results demonstrate how template-based approaches now effectively take over for our previous single variable approaches and produce quality intervention targets in larger networks requiring sampled state spaces.

DDI-CPI, a server that predicts drug-drug interactions through implementing the chemical-protein interactome

Drug–drug interactions (DDIs) may cause serious side-effects that draw great attention from both academia and industry. Since some DDIs are mediated by unexpected drug–human protein interactions, it is reasonable to analyze the chemical–protein interactome (CPI) profiles of the drugs to predict their DDIs. Here we introduce the DDI-CPI server, which can make real-time DDI predictions based only on molecular structure. When the user submits a molecule, the server will dock user's molecule across 611 human proteins, generating a CPI profile that can be used as a feature vector for the pre-constructed prediction model. It can suggest potential DDIs between the user's molecule and our library of 2515 drug molecules. In cross-validation and independent validation, the server achieved an AUC greater than 0.85. Additionally, by investigating the CPI profiles of predicted DDI, users can explore the PK/PD proteins that might be involved in a particular DDI. A 3D visualization of the drug-protein interaction will be provided as well. The DDI-CPI is freely accessible at http://cpi.bio-x.cn/ddi/.

Pharmacointeraction network models predict unknown drug-drug interactions

Drug-drug interactions (DDIs) can lead to serious and potentially lethal adverse events. In recent years, several drugs have been withdrawn from the market due to interaction-related adverse events (AEs). Current methods for detecting DDIs rely on the accumulation of sufficient clinical evidence in the post-market stage – a lengthy process that often takes years, during which time numerous patients may suffer from the adverse effects of the DDI. Detection methods are further hindered by the extremely large combinatoric space of possible drug-drug-AE combinations. There is therefore a practical need for predictive tools that can identify potential DDIs years in advance, enabling drug safety professionals to better prioritize their limited investigative resources and take appropriate regulatory action. To meet this need, we describe Predictive Pharmacointeraction Networks (PPINs) – a novel approach that predicts unknown DDIs by exploiting the network structure of all known DDIs, together with other intrinsic and taxonomic properties of drugs and AEs. We constructed an 856-drug DDI network from a 2009 snapshot of a widely-used drug safety database, and used it to develop PPIN models for predicting future DDIs. We compared the DDIs predicted based solely on these 2009 data, with newly reported DDIs that appeared in a 2012 snapshot of the same database. Using a standard multivariate approach to combine predictors, the PPIN model achieved an AUROC (area under the receiver operating characteristic curve) of 0.81 with a sensitivity of 48% given a specificity of 90%. An analysis of DDIs by severity level revealed that the model was most effective for predicting “contraindicated” DDIs (AUROC = 0.92) and less effective for “minor” DDIs (AUROC = 0.63). These results indicate that network based methods can be useful for predicting unknown drug-drug interactions.

Assessment of hydroxylated metabolites of polychlorinated biphenyls as potential xenoestrogens: a QSAR comparative analysis∗

Alternative methods, including quantitative structure-activity relationships (QSAR), are being used increasingly when appropriate data for toxicity evaluation of chemicals are not available. Approximately 40 mono-hydroxylated polychlorinated biphenyls (OH-PCBs) have been identified in humans. They represent a health and environmental concern because some of them have been shown to have agonist or antagonist interactions with human hormone receptors. This could lead to modulation of steroid hormone receptor pathways and endocrine system disruption. We performed QSAR analyses using available estrogenic activity (human estrogen receptor ER alpha) data for 71 OH-PCBs. The modelling was performed using multiple molecular descriptors including electronic, molecular, constitutional, topological, and geometrical endpoints. Multiple linear regressions and recursive partitioning were used to best fit descriptors. The results show that the position of the hydroxyl substitution, polarizability, and meta adjacent un-substituted carbon pairs at the phenolic ring contribute towards greater estrogenic activity for these chemicals. These comparative QSAR models may be used for predictive toxicity, and identification of health consequences of PCB metabolites that lack empirical data. Such information will help prioritize such molecules for additional testing, guide future basic laboratory research studies, and help the health/risk assessment community understand the complex nature of chemical mixtures.

INDI: a computational framework for inferring drug interactions and their associated recommendations

INDI is a similarity-based drug–drug interaction prediction method that can infer both pharmacokinetic and pharmacodynamic interactions, as well as their severity levels. Both known and predicted drug interactions are found to be prevalent in clinical practice.

INDI is a similarity-based drug–drug interaction prediction method, capable of handling both pharmacokinetic and pharmacodynamic interactions.

INDI predicts the severity of the interaction and the Cytochrome P450 isozyme involved in pharmacokinetic interactions.

We show the prevalence of known and predicted drug interactions in drug adverse reports and in chronic medications taken by hospitalized patients.

Inferring drug–drug interactions (DDIs) is an essential step in drug development and drug administration. Most computational inference methods focus on modeling drug pharmacokinetics, aiming at interactions that result from a common metabolizing enzyme (CYP). Here, we introduce a novel prediction method, INDI (INferring Drug Interactions), allowing the inference of both pharmacokinetic, CYP-related DDIs (along with their associated CYPs) and pharmacodynamic, non-CYP associated ones. On cross validation, it obtains high specificity and sensitivity levels (AUC (area under the receiver-operating characteristic curve)⩾0.93). In application to the FDA adverse event reporting system, 53% of the drug events could potentially be connected to known (41%) or predicted (12%) DDIs. Additionally, INDI predicts the severity level of each DDI upon co-administration of the involved drugs, suggesting that severe interactions are abundant in the clinical practice. Examining regularly taken medications by hospitalized patients, 18% of the patients receive known or predicted severely interacting drugs and are hospitalized more frequently. Access to INDI and its predictions is provided via a web tool at http://www.cs.tau.ac.il/∼bnet/software/INDI, facilitating the inference and exploration of drug interactions and providing important leads for physicians and pharmaceutical companies alike.

Predicting the clinical relevance of drug interactions from pre-approval studies

Drug interactions (DIs) may result in adverse drug events that could be prevented, but in many cases the available information on potential DIs is not easily transferable to clinical practice. The majority of studies date from preclinical or premarketing phases, using animals or human-derived sources that may not accurately reflect the growing clinical complexity of high-risk populations, such as the elderly, women, children, patients with chronic disease, polytherapy and impaired organ functions. Thus, at the time of approval of a new drug the information in the summary of product characteristics refers to potential DIs, but lacks specific management recommendations and is of limited clinical utility. Therefore, we set out to review in vitro and in vivo methods to predict and quantify potential DIs, to see whether these studies could help the physician tackle daily problems of the assessment and choice of combined drug therapies, and to propose, from a clinical point of view, how premarketing studies could be improved so as to help the physician at the patient's bedside. Preclinical and premarketing study design needs to be improved to make information easily accessible and clinically transferable. Studies should also take into account appropriate sample size, duration, co-morbidity, number of coadministered drugs, within- and between-subject variability, specific at-risk populations and/or drugs with a relatively narrow therapeutic window, and clinical endpoints. After premarketing development in situations where there is potential high risk of serious adverse events, specific phase IV studies (and/or active pharmacovigilance studies) should be required to monitor and quantitatively assess their clinical impact.

Understanding CYP2D6 interactions

Owing to the polymorphic nature of CYP2D6, clinically significant issues can arise when drugs rely on that enzyme either for clearance, or metabolism to an active metabolite. Available screening methods to determine if the compound is likely to cause drug-drug interactions, or is likely to be a victim of inhibition of CYP2D6 by other compounds will be described. Computational models and examples will be given on strategies to design out the CYP2D6 liabilities for both heme-binding compounds and non-heme-binding compounds.

In silico log P prediction for a large data set with support vector machines, radial basis neural networks and multiple linear regression

Oil/water partition coefficient (log P) is one of the key points for lead compound to be drug. In silico log P models based solely on chemical structures have become an important part of modern drug discovery. Here, we report support vector machines, radial basis function neural networks, and multiple linear regression methods to investigate the correlation between partition coefficient and physico-chemical descriptors for a large data set of compounds. The correlation coefficient r(2) between experimental and predicted log P for training and test sets by support vector machines, radial basis function neural networks, and multiple linear regression is 0.92, 0.90, and 0.88, respectively. The results show that non-linear support vector machines derives statistical models that have better prediction ability than those of radial basis function neural networks and multiple linear regression methods. This indicates that support vector machines can be used as an alternative modeling tool for quantitative structure-property/activity relationships studies.

Functional evolution of the pregnane X receptor

The pregnane X receptor (PXR; NR1I2) is a nuclear hormone receptor (NR) that transcriptionally regulates genes encoding transporters and drug-metabolising enzymes in the liver and intestine. PXR activation leads to enhanced metabolism and elimination of xenobiotics and endogenous compounds such as hormones and bile salts. Relative to other vertebrate NRs, PXR has the broadest specificity for ligand activators by virtue of a large, flexible ligand-binding cavity. In addition, PXR has the most extensive sequence diversity across vertebrate species in the ligand-binding domain of any NR, with significant pharmacological differences between human and rodent PXRs, and especially marked divergence between mammalian and nonmammalian PXRs. The unusual properties of PXR complicate the use of in silico and animal models to predict in vivo human PXR pharmacology. Research into the evolutionary history of the PXR gene has also provided insight into the function of PXR in humans and other animals.

Simulation and prediction of in vivo drug metabolism in human populations from in vitro data

The perceived failure of new drug development has been blamed on deficiencies in in vivo studies of drug efficacy and safety. Prior simulation of the potential exposure of different individuals to a given dose might help to improve the design of such studies. This should also help researchers to focus on the characteristics of individuals who present with extreme reactions to therapy. An effort to build virtual populations using extensive demographic, physiological, genomic and in vitro biochemical data to simulate and predict drug disposition from routinely collected in vitro data is outlined.

Mechanism-based inhibition of cytochrome P450 3A4 by therapeutic drugs

Consistent with its highest abundance in humans, cytochrome P450 (CYP) 3A is responsible for the metabolism of about 60% of currently known drugs. However, this unusual low substrate specificity also makes CYP3A4 susceptible to reversible or irreversible inhibition by a variety of drugs. Mechanism-based inhibition of CYP3A4 is characterised by nicotinamide adenine dinucleotide phosphate hydrogen (NADPH)-, time- and concentration-dependent enzyme inactivation, occurring when some drugs are converted by CYP isoenzymes to reactive metabolites capable of irreversibly binding covalently to CYP3A4. Approaches using in vitro, in silico and in vivo models can be used to study CYP3A4 inactivation by drugs. Human liver microsomes are always used to estimate inactivation kinetic parameters including the concentration required for half-maximal inactivation (K(I)) and the maximal rate of inactivation at saturation (k(inact)). Clinically important mechanism-based CYP3A4 inhibitors include antibacterials (e.g. clarithromycin, erythromycin and isoniazid), anticancer agents (e.g. tamoxifen and irinotecan), anti-HIV agents (e.g. ritonavir and delavirdine), antihypertensives (e.g. dihydralazine, verapamil and diltiazem), sex steroids and their receptor modulators (e.g. gestodene and raloxifene), and several herbal constituents (e.g. bergamottin and glabridin). Drugs inactivating CYP3A4 often possess several common moieties such as a tertiary amine function, furan ring, and acetylene function. It appears that the chemical properties of a drug critical to CYP3A4 inactivation include formation of reactive metabolites by CYP isoenzymes, preponderance of CYP inducers and P-glycoprotein (P-gp) substrate, and occurrence of clinically significant pharmacokinetic interactions with coadministered drugs. Compared with reversible inhibition of CYP3A4, mechanism-based inhibition of CYP3A4 more frequently cause pharmacokinetic-pharmacodynamic drug-drug interactions, as the inactivated CYP3A4 has to be replaced by newly synthesised CYP3A4 protein. The resultant drug interactions may lead to adverse drug effects, including some fatal events. For example, when aforementioned CYP3A4 inhibitors are coadministered with terfenadine, cisapride or astemizole (all CYP3A4 substrates), torsades de pointes (a life-threatening ventricular arrhythmia associated with QT prolongation) may occur.However, predicting drug-drug interactions involving CYP3A4 inactivation is difficult, since the clinical outcomes depend on a number of factors that are associated with drugs and patients. The apparent pharmacokinetic effect of a mechanism-based inhibitor of CYP3A4 would be a function of its K(I), k(inact) and partition ratio and the zero-order synthesis rate of new or replacement enzyme. The inactivators for CYP3A4 can be inducers and P-gp substrates/inhibitors, confounding in vitro-in vivo extrapolation. The clinical significance of CYP3A inhibition for drug safety and efficacy warrants closer understanding of the mechanisms for each inhibitor. Furthermore, such inactivation may be exploited for therapeutic gain in certain circumstances.

Food-drug interactions via human cytochrome P450 3A (CYP3A)

Food-drug interactions have been reported to occur in various systems in the body. The causes of these interactions are mainly divided into pharmacodynamic and pharmacokinetic processes. Among these processes, drug metabolism plays a crucial role in drug interactions. Metabolic food-drug interactions occur when a certain food alters the activity of a drug-metabolizing enzyme, leading to a modulation of the pharmacokinetics of drugs metabolized by the enzyme. A variety of interactions have been documented so far. Foods consisting of complex chemical mixtures, such as fruits, alcoholic beverages, teas, and herbs, possess the ability to inhibit or induce the activity of drug-metabolizing enzymes. According to results obtained thus far, cytochrome P450 3A4 (CYP3A4) appears to be a key enzyme in food-drug interactions. For example, interactions of grapefruit juice with felodipine and cyclosporine, red wine with cyclosporine, and St John's wort with various medicines including cyclosporine, have been demonstrated. The results indicate the requirement of dosage adjustment to maintain drug concentrations within their therapeutic windows. The CYP3A4-related interaction by food components may be related to the high level of expression of CYP3A4 in the small intestine, as well as its broad substrate specificity, as CYP3A4 is responsible for the metabolism of more than 50% of clinical pharmaceuticals. This review article summarizes the findings obtained to date concerning food-drug interactions and their clinical implications. It seems likely that more information regarding such interactions will accumulate in the future, and awareness is necessary for achieving optimal drug therapy.

Structure-based methods for the prediction of the dominant P450 enzyme in human drug biotransformation: consideration of CYP3A4, CYP2C9, CYP2D6

Metabolic drug-drug interactions are receiving more and more attention from the in silico community. Early prediction of such interactions would not only improve drug safety but also contribute to make drug design more predictable and rational. The aim of this study was to build a simple and interpretable model for the determination of the P450 enzyme predominantly responsible for a drug's metabolism. The P450 enzymes taken into consideration were CYP3A4, CYP2D6 and CYP2C9. Physico-chemical descriptors and structural descriptors for 96 currently marketed drugs were submitted to statistical analysis using the formal inference-based recursive modelling (FIRM) method, a form of recursive partitioning. Generally accepted knowledge on metabolism by these enzymes was also used to construct a hierarchical decision tree. Robust methods of variable selection using recursive partitioning were utilised. The descriptive ability of the resulting hierarchical model is very satisfactory, with 94% of the compounds correctly classified.

[High throughput screening of pharmacokinetics and metabolism in drug discovery (III)--investigation on in- silico model for membrane permeability and CYP1A2 inhibition]

Pharmacokinetic and metabolic screening plays an important role in the optimization of a lead compound in drug discovery. Since those screening methods are time-consuming and labor intensive, in silico models would be effective to select compounds and guide derivatization prior to the screening. We investigated in silico models for permeability in Caco-2 cells, brain distribution and cytochrome P450 (CYP) inhibition using molecular weight, lipophilicity (clog D(7.4)), polar surface area (PSA), and number of rotatable bonds (RB). A variety of test compounds was selected from different Caco-2 assay projects. The permeability determined exhibited a good correlation with a combination of PSA and clog D(7.4) rather than with PSA alone. In the brain distribution, PSA, in addition to lipophilicity, was one of the determinant parameters, and compounds were significantly distributed to the brain in rats with the decrease in the PSA value. When this approach was adapted to CYP1A2 inhibition in the fluorometric assay, the inhibitory potential for two plane core structures was successfully predicted by utilizing number of RB, PSA, and clog D(7.4). In particular, an increase in the number of RB weakened the inhibitory potential due to a loss of the plane structures. These results suggest that the PSA and RB are key parameters to design chemical structures in terms of the improvement of both membrane permeability in the brain and gastrointestine and CYP1A2 inhibition, respectively.

Active-site characteristics of CYP2C19 and CYP2C9 probed with hydantoin and barbiturate inhibitors

Three series of N-3 alkyl substituted phenytoin, nirvanol, and barbiturate derivatives were synthesized and their inhibitor potencies were tested against recombinant CYP2C19 and CYP2C9 to probe the interaction of these ligands with the active sites of these enzymes. All compounds were found to be competitive inhibitors of both enzymes, although the degree of inhibitory potency was generally much greater towards CYP2C19. Inhibitor stereochemistry did not markedly influence K(i) towards CYP2C9, and log P adequately predicted inhibitor potency for this enzyme. In contrast, stereochemistry was an important factor in determining inhibitor potency towards CYP2C19. (S)-(+)-N-3-Benzylnirvanol and (R)-(-)-N-3-benzylphenobarbital emerged as the most potent and selective CYP2C19 inhibitors, with K(i) values of < 250nM--at least two orders of magnitude greater inhibitor potency than towards CYP2C9. Both inhibitors were metabolized preferentially at their C-5 phenyl substituents, indicating that CYP2C19 prefers to orient the N-3 substituents away from the active oxygen species. These features were incorporated into expanded CoMFA models for CYP2C9, and a new, validated CoMFA model for CYP2C19.

Gastrointestinal-tract models and techniques for use in safety pharmacology

The gastrointestinal tract (GI tract), which extracts nutrients, electrolytes, minerals, and water, is prone to injury as a result of oral drug administration. Clinical assessment of the GI tract is often limited to measurements of transit time and observations of vomiting or diarrhoea, despite the existence of methods and techniques capable of assessing specific changes in GI function at the membrane, cell, and whole animal levels. Membrane studies, record the uptake of solutes, and electrolyte transport, assessing the affects of compounds on transepithelial GI transport and flux. Such methods lend themselves to permeability, immunohistochemistry, morphology, and molecular biology techniques. Isolated cells from the GI tract or cultured cell lines provide knowledge of regulation and function at a cellular level, whilst motility patterns, taken in vivo or from biopsies, provide information at a more integrated level. In anesthetised animals, ligated segments of the intestine can be infused with test compounds, providing information about absorptive and secretory processes important for the treatment of diarrhoea. Computer simulations and modelling are used to predict the disposition of a chemical and its metabolite and can, to some extent, replace animal testing, thereby reducing development costs. Indeed, software programs can be used to simulate the dissolution, absorption, distribution, metabolism, and excretion (ADME) properties of potential drugs in the human GI tract. Finally, advances in the field of imaging, combined with endoscopy, have resulted in a wireless capsule, allowing the inspection of the GI tract anatomy and pathology without surgical intervention. It is concluded that the field of safety pharmacology could rapidly, cheaply, and routinely incorporate membrane, isolated tissue, and endoscopy techniques for GI tract testing of drugs.

ADME-Tox in drug discovery: integration of experimental and computational technologies

Over the past ten years, in vitro experimental tools to characterize ADME-Tox profiles of compounds have been applied in early stages of the drug discovery process to increase the success rate of discovery programmes and to progress better candidates into drug development. Application of in silico ADME-Tox models has further enhanced discovery support, enabling virtual screening of compounds and thus, application of ADME-Tox at every stage of the discovery process. Ultimately, effective and efficient ADME-Tox support of discovery will depend on a complementary and synergistic use of experimental and in silico ADME-Tox.

A multiple in silico program approach for the prediction of mutagenicity from chemical structure

We have conducted an evaluation of three of the most widely used commercial toxicity prediction programs, Toxicity Prediction by Komputer Assisted Technology (TOPKAT), Deductive Estimation of Risk from Existing Knowledge (DEREK) for Windows (DfW) and CASETOX. The three programs were evaluated for their ability to predict Ames test mutagenicity using 520 proprietary drug candidate (Test set 1) and 94 commercial (Test set 2) compounds. The study demonstrates that these three commercially available programs are useful, with limitations in their ability to predict mutagenicity over a wide range of chemical space, i.e. global predictivity. Individually, each of the programs performed at an acceptable level for overall accuracy, i.e. the ability to predict the correct outcome. However, analysis of the predictions indicates that the overall accuracy figure is heavily weighted by the ability of the programs to correctly predict non-mutagens, whereas none of the programs individually performed well in the prediction of novel mutagenic structures, i.e. Ames positive compounds. The performance of these programs' in predicting Ames positive mutagens appeared to be independent of the chemical utility of the compound, i.e. industrial, agricultural or pharmaceutical. The combination of program predictions provided some improvement in overall accuracy, sensitivity and specificity.

ADMET in silico modelling: towards prediction paradise?

Following studies in the late 1990s that indicated that poor pharmacokinetics and toxicity were important causes of costly late-stage failures in drug development, it has become widely appreciated that these areas should be considered as early as possible in the drug discovery process. However, in recent years, combinatorial chemistry and high-throughput screening have significantly increased the number of compounds for which early data on absorption, distribution, metabolism, excretion (ADME) and toxicity (T) are needed, which has in turn driven the development of a variety of medium and high-throughput in vitro ADMET screens. Here, we describe how in silico approaches will further increase our ability to predict and model the most relevant pharmacokinetic, metabolic and toxicity endpoints, thereby accelerating the drug discovery process.

Interactions of herbs with cytochrome P450

A resurgence in the use of medical herbs in the Western world, and the co-use of modern and traditional therapies is becoming more common. Thus there is the potential for both pharmacokinetic and pharmacodynamic herb-drug interactions. For example, systems such as the cytochrome P450 (CYP) may be particularly vulnerable to modulation by the multiple active constituents of herbs, as it is well known that the CYPs are subject to induction and inhibition by exposure to a wide variety of xenobiotics. Using in vitro, in silico, and in vivo approaches, many herbs and natural compounds isolated from herbs have been identified as substrates, inhibitors, and/or inducers of various CYP enzymes. For example, St. John's wort is a potent inducer of CYP3A4, which is mediated by activating the orphan pregnane X receptor. It also contains ingredients that inhibit CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4. Many other common medicinal herbs also exhibited inducing or inhibiting effects on the CYP system, with the latter being competitive, noncompetitive, or mechanism-based. It appears that the regulation of CYPs by herbal products complex, depending on the herb type, their administration dose and route, the target organ and species. Due to the difficulties in identifying the active constituents responsible for the modulation of CYP enzymes, prediction of herb-drug metabolic interactions is difficult. However, herb-CYP interactions may have important clinical and toxicological consequences. For example, induction of CYP3A4 by St. John's wort may partly provide an explanation for the enhanced plasma clearance of a number of drugs, such as cyclosporine and innadivir, which are known substrates of CYP3A4, although other mechanisms including modulation of gastric absorption and drug transporters cannot be ruled out. In contrast, many organosulfur compounds, such as diallyl sulfide from garlic, are potent inhibitors of CYP2E1; this may provide an explanation for garlic's chemoproventive effects, as many mutagens require activation by CYP2E1. Therefore, known or potential herb-CYP interactions exist, and further studies on their clinical and toxicological roles are warranted. Given that increasing numbers of people are exposed to a number of herbal preparations that contain many constituents with potential of CYP modulation, high-throughput screening assays should be developed to explore herb-CYP interactions.

In silico prediction of ADME and pharmacokinetics. Report of an expert meeting organised by COST B15

The computational approach is one of the newest and fastest developing techniques in pharmacokinetics, ADME (absorption, distribution, metabolism, excretion) evaluation, drug discovery and toxicity. However, to date, the software packages devoted to ADME prediction, especially of metabolism, have not yet been adequately validated and still require improvements to be effective. Most are 'open' systems, under constant evolution and able to incorporate rapidly, and often easily, new information from user or developer databases. Quantitative in silico predictions are now possible for several pharmacokinetic (PK) parameters, particularly absorption and distribution. The emerging consensus is that the predictions are no worse than those made using in vitro tests, with the decisive advantage that much less investment in technology, resources and time is needed. In addition, and of critical importance, it is possible to screen virtual compounds. Some packages are able to handle thousands of molecules in a few hours. However, common experience shows that, in part at least for essentially irrational reasons, there is currently a lack of confidence in these approaches. An effort should be made by the software producers towards more transparency, in order to improve the confidence of their consumers. It seems highly probable that in silico approaches will evolve rapidly, as did in vitro methods during the last decade. Past experience with the latter should be helpful in avoiding repetition of similar errors and in taking the necessary steps to ensure effective implementation. A general concern is the lack of access to the large amounts of data on compounds no longer in development, but still kept secret by the pharmaceutical industry. Controlled access to these data could be particularly helpful in validating new in silico approaches.

In vitro tests for predicting drug-drug interactions: the need for validated procedures

Over the past decade, the prediction of drug-drug interactions from in vitro studies has become a rapidly expanding field of research. Numerous papers and excellent review articles (Bertz & Granneman 1997; Ito et al. 1998a & b; Lin 2000; Bachmann & Ghosh 2001; Ekins & Wrighton 2001; Weaver 2001) have been published in this area. Yet like any new and fast-growing subject, this one has been developing with some confusion and without any real, efficient organisation. Depending on the drug tested, the models and extrapolation parameters used, etc., results and conclusions may vary widely from study to study (von Moltke et al. 1998; Weaver 2001). Several authors have called for validation of these procedures (Rodrigues et al. 2001; Kummar & Surapaneni 2001; Pelkonen et al. 2001a & b; Kremers 2002), and regulatory authorities intend to require better traceability and reliability (FDA & EMEA guidelines). A systematic and reliable approach is needed also to allow such protocols to be incorporated into early screening for potential drugs and new chemical entities. There is certainly a great need to standardise these studies and to verify their conclusions, but is true validation possible in this field? The main purpose of the present paper is to discuss this issue and to examine what is possible and what is needed to improve the quality of predictions made from in vitro experiments.

Current practices in generation of small molecule new leads

The current drug discovery processes in many pharmaceutical companies require large and growing collections of high quality lead structures for use in high throughput screening assays. Collections of small molecules with diverse structures and "drug-like" properties have, in the past, been acquired by several means: by archive of previous internal lead optimization efforts, by purchase from compound vendors, and by union of separate collections following company mergers. More recently, many drug discovery companies have established dedicated efforts to effect synthesis by internal and/or outsourcing efforts of targeted compound libraries for new lead generation. Although high throughput/combinatorial chemistry is an important component in the process of new lead generation, the selection of library designs for synthesis and the subsequent design of library members has evolved to a new level of challenge and importance. The potential benefits of screening multiple small molecule compound library designs against multiple biological targets offers substantial opportunity to discover new lead structures. Subsequent optimization of such compounds is often accelerated because of the structure-activity relationship (SAR) information encoded in these lead generation libraries. Lead optimization is often facilitated due to the ready applicability of high-throughput chemistry (HTC) methods for follow-up synthesis. Some of the strategies, trends, and critical issues central to the success of lead generation processes are discussed below.

Pharmacophore modeling of cytochromes P450

Understanding the binding of ligands in the active site of a membrane-bound protein is difficult in the absence of a crystal structure. When these proteins are the enzymes involved in drug metabolism, it leaves little option but to use site-directed mutagenesis and in vitro studies to provide critical information relating to determinants of binding affinity. Pharmacophore models and three-dimensional quantitative structure-activity relationships have been used either alone or in combination with protein homology models to provide this information for cytochrome P450s. At present, their application has been directed to the major enzymes but this may escalate in future as more in vitro data are generated for other P450s. The following review outlines the methodologies and models as well as future prospects for applying these technologies to P450s in the hope that future drugs will be selected with increased metabolic stability and fewer incidences of undesirable drug-drug interactions.