Prediction Of Drug-drug Interactions Research Articles

Selection of an Optimal In Vitro Model to Assess P-gp Inhibition: Comparison of Vesicular and Bidirectional Transcellular Transport Inhibition Assays

The multidrug resistance protein 1 (MDR1) P-glycoprotein (P-gp) is a clinically important transporter. In vitro P-gp inhibition assays have been routinely conducted to predict the potential for clinical drug-drug interactions (DDIs) mediated by P-gp. However, high inter- laboratory and inter-system variability of P-gp IC50 data limits accurate prediction of DDIs using static models and decision criteria recommended by regulatory agencies. In this study, we calibrated two in vitro P-gp inhibition models: vesicular uptake of N-methyl-quinidine (NMQ) in MDR1 vesicles and bidirectional transport (BDT) of digoxin in Lilly Laboratories Cell Porcine Kidney 1 cells overexpressing MDR1 (LLC-MDR1) using a total of 48 P-gp inhibitor and non-inhibitor drugs, and digoxin DDI data from 70 clinical studies. Refined thresholds were derived using receiver operating characteristic (ROC) analysis and their predictive performance was compared with the decision frameworks proposed by regulatory agencies and selected reference. Furthermore, the impact of various IC50 calculation methods and non-specific binding of drugs on DDI prediction was evaluated. Our studies suggest that the concentration of inhibitor based on highest approved dose dissolved in 250 ml divided by IC50(I2/IC50) is sufficient to predict P-gp related intestinal DDIs. IC50 obtained from vesicular inhibition assay with a refined threshold of I2/IC50 {greater than or equal to} 25.9 provides comparable predictive power than those measured by net secretory flux and efflux ratio in LLC-MDR1 cells. We therefore recommend vesicular P-gp inhibition as our preferred method given its simplicity, lower variability, higher assay throughput, and more direct estimation of in vitro kinetic parameters than BDT assay. Significance Statement We have conducted comprehensive calibration of two in vitro P-gp inhibition models: uptake in MDR1 vesicles and bidirectional transport in LLC-MDR1 cell monolayers to predict DDIs. Our studies suggest that IC50s obtained from vesicular inhibition with a refined threshold of I2/IC50 ≥ 25.9 provide comparable predictive power than those in LLC-MDR1 cells. We therefore recommend vesicular P-gp inhibition as preferred method given its simplicity, lower variability, higher assay throughput, and more direct estimation of in vitro kinetic parameters.

Physiologically Based Pharmacokinetic (PBPK) Modeling of Clopidogrel and Its Four Relevant Metabolites for CYP2B6, CYP2C8, CYP2C19, and CYP3A4 Drug-Drug-Gene Interaction Predictions.

The antiplatelet agent clopidogrel is listed by the FDA as a strong clinical index inhibitor of cytochrome P450 (CYP) 2C8 and weak clinical inhibitor of CYP2B6. Moreover, clopidogrel is a substrate of—among others—CYP2C19 and CYP3A4. This work presents the development of a whole-body physiologically based pharmacokinetic (PBPK) model of clopidogrel including the relevant metabolites, clopidogrel carboxylic acid, clopidogrel acyl glucuronide, 2-oxo-clopidogrel, and the active thiol metabolite, with subsequent application for drug–gene interaction (DGI) and drug–drug interaction (DDI) predictions. Model building was performed in PK-Sim® using 66 plasma concentration-time profiles of clopidogrel and its metabolites. The comprehensive parent-metabolite model covers biotransformation via carboxylesterase (CES) 1, CES2, CYP2C19, CYP3A4, and uridine 5′-diphospho-glucuronosyltransferase 2B7. Moreover, CYP2C19 was incorporated for normal, intermediate, and poor metabolizer phenotypes. Good predictive performance of the model was demonstrated for the DGI involving CYP2C19, with 17/19 predicted DGI AUClast and 19/19 predicted DGI Cmax ratios within 2-fold of their observed values. Furthermore, DDIs involving bupropion, omeprazole, montelukast, pioglitazone, repaglinide, and rifampicin showed 13/13 predicted DDI AUClast and 13/13 predicted DDI Cmax ratios within 2-fold of their observed ratios. After publication, the model will be made publicly accessible in the Open Systems Pharmacology repository.

Effect of Cyclosporin A and Impact of Dose Staggering on OATP1B1/1B3 Endogenous Substrates and Drug Probes for Assessing Clinical Drug Interactions.

This study was designed to assess the quantitative performance of endogenous biomarkers for organic anion transporting polypeptide (OATP) 1B1/1B3‐mediated drug‐drug interactions (DDIs). Ten healthy volunteers orally received OATP1B1/1B3 probe cocktail (0.2 mg pitavastatin, 1 mg rosuvastatin, and 2 mg valsartan) and an oral dose of cyclosporin A (CysA, 20 mg and 75 mg) separated by a 1‐hour interval (20 mg (−1 hour), and 75 mg (−1 hour)). CysA 75 mg was also given with a 3‐hour interval (75 mg (−3 hours)) to examine the persistence of OATP1B1/1B3 inhibition. The area under the plasma concentration‐time curve ratios (AUCRs) were 1.63, 3.46, and 2.38 (pitavastatin), 1.39, 2.16, and 1.81 (rosuvastatin), and 1.42, 1.77, and 1.85 (valsartan), at 20 mg, 75 mg (−1 hour) and 75 mg (−3 hours) of CysA, respectively. CysA effect on OATP1B1/1B3 was unlikely to persist at the dose examined. Among 26 putative OATP1B1/1B3 biomarkers evaluated, AUCR and maximum concentration ratio (C maxR) of CP‐I showed the highest Pearson’s correlation coefficient with CysA AUC (0.94 and 0.93, respectively). Correlation between AUCR of pitavastatin, and C maxR or AUCR of CP‐I were consistent between this study and our previous study using rifampicin as an OATP1B1/1B3 inhibitor. Nonlinear regression analysis of AUCR−1 of pitavastatin and CP‐I against CysA C max yielded K i,OATP1B1/1B3,app (109 ± 35 and 176 ± 42 nM, respectively), similar to the Ki ,OATP1B1/1B3 estimated by our physiologically‐based pharmacokinetic model analysis described previously (107 nM). The endogenous OATP1B1/1B3 biomarkers, particularly C maxR and AUCR of CP‐I, corroborates OATP1B1/1B3 inhibition and yields valuable information that improve accurate DDI predictions in drug development, and enhance our understanding of interindividual variability in the magnitude of DDIs.

Quantitative prediction of OATP1B-mediated drug–drug interactions using endogenous biomarker coproporphyrin I

Evaluation of the organic anion transporting polypeptide (OATP) 1B-mediated drug–drug interaction (DDI) potential is important for drug development. The focus of this study was coproporphyrin I (CP-I), an endogenous OATP1B biomarker. We investigated a new approach to OATP1B-mediated DDI prediction based on the mechanistic static pharmacokinetics (MSPK) model. The ratio of the area under the plasma concentration–time curve (AUCR) with and without co-administration of rifampicin (a typical OATP1B inhibitor) was found for CP-I and OATP1B substrate, respectively, and was then used to derive the correlation curve equation. The AUCR with and without co-administration of another OATP1B inhibitor than rifampicin was then predicted for the OATP1B substrates by substituting the AUCR of CP-I in the correlation curve equation to verify the predictability of the AUCR of the OATP1B substrates. The derived correlation curve equation between CP-I and the OATP1B substrates of the AUCRs with and without co-administration of rifampicin matched the observed AUCRs well. Regarding pitavastatin, rosuvastatin, and pravastatin, 92.9% of the predicted AUCR values were within a two-fold range of the observed values, indicating that this approach may be a good way to quantitatively predict DDI potential.

BioDKG-DDI: predicting drug-drug interactions based on drug knowledge graph fusing biochemical information.

The way of co-administration of drugs is a sensible strategy for treating complex diseases efficiently. Because of existing massive unknown interactions among drugs, predicting potential adverse drug-drug interactions (DDIs) accurately is promotive to prevent unanticipated interactions, which may cause significant harm to patients. Currently, numerous computational studies are focusing on potential DDIs prediction on account of traditional experiments in wet lab being time-consuming, labor-consuming, costly and inaccurate. These approaches performed well; however, many approaches did not consider multi-scale features and have the limitation that they cannot predict interactions among novel drugs. In this paper, we proposed a model of BioDKG-DDI, which integrates multi-feature with biochemical information to predict potential DDIs through an attention machine with superior performance. Molecular structure features, representation of drug global association using drug knowledge graph (DKG) and drug functional similarity features are fused by attention machine and predicted through deep neural network. A novel negative selecting method is proposed to certify the robustness and stability of our method. Then, three datasets with different sizes are used to test BioDKG-DDI. Furthermore, the comparison experiments and case studies can demonstrate the reliability of our method. Upon our finding, BioDKG-DDI is a robust, yet simple method and can be used as a benefic supplement to the experimental process.

Predicting Potential Drug-Drug-Gene Interactions in a Population of Individuals Utilizing a Community-Based Pharmacy

Adverse drug reactions (ADRs) are associated with increased morbidity, mortality, and resource utilization. Drug interactions (DDIs) are among the most common causes of ADRs, and estimates have cited that up to 22% of patients take interacting medications. DDIs are often due to the propensity for agents to induce or inhibit enzymes responsible for the metabolism of concomitantly administered drugs. However, this phenomenon is further complicated by genetic variants of such enzymes. The aim of this study is to quantify and describe potential drug-drug, drug-gene, and drug-drug-gene interactions in a community-based patient population. A regional pharmacy with retail outlets in Arkansas provided deidentified prescription data from March 2020 for 4761 individuals. Drug-drug and drug-drug-gene interactions were assessed utilizing the logic incorporated into GenMedPro, a commercially available digital gene-drug interaction software program that incorporates variants of 9 pharmacokinetic (PK) and 2 pharmacodynamic (PD) genes to evaluate DDIs and drug-gene interactions. The data were first assessed for composite drug-drug interaction risk, and each individual was stratified to a risk category using the logic incorporated in GenMedPro. To calculate the frequency of potential drug-gene interactions, genotypes were imputed and allocated to the cohort according to each gene's frequency in the general population. Potential genotypes were randomly allocated to the population 100 times in a Monte Carlo simulation. Potential drug-drug, gene-drug, or gene-drug-drug interaction risk was characterized as minor, moderate, or major. Based on prescription data only, the probability of a DDI of any impact (mild, moderate, or major) was 26% [95% CI: 0.248-0.272] in the population. This probability increased to 49.6% [95% CI: 0.484-0.507] when simulated genetic polymorphisms were additionally assessed. When assessing only major impact interactions, there was a 7.8% [95% CI: 0.070-0.085] probability of drug-drug interactions and 10.1% [95% CI: 0.095-0.108] probability with the addition of genetic contributions. The probability of drug-drug-gene interactions of any impact was correlated with the number of prescribed medications, with an approximate probability of 77%, 85%, and 94% in patients prescribed 5, 6, or 7+ medications, respectively. When stratified by specific drug class, antidepressants (19.5%), antiemetics (21.4%), analgesics (16%), antipsychotics (15.6%), and antiparasitics (49.7%) had the highest probability of major drug-drug-gene interaction. In a community-based population of outpatients, the probability of drug-drug interaction risk increases when genetic polymorphisms are attributed to the population. These data suggest that pharmacogenetic testing may be useful in predicting drug interactions, drug-gene interactions, and severity of interactions when proactively evaluating patient medication profiles. Genomind, Inc.

In Vitro Assessment of Transporter Mediated Perpetrator DDIs for Several Hepatitis C Virus Direct-Acting Antiviral Drugs and Prediction of DDIs with Statins Using Static Models.

Inhibitory effects of asunaprevir, daclatasvir, grazoprevir, paritaprevir, simeprevir, and voxilaprevir, direct-acting antiviral (DAA) drugs for the treatment of chronic hepatitis C virus (HCV) infection, were evaluated in vitro against a range of clinically important drug transporters. In vitro inhibition studies were conducted using transporter transfected cells and membrane vesicles. The risk of clinical drug-drug interactions (DDIs) was assessed using simplified static models recommended by regulatory agencies. Furthermore, we refined and developed static models to predict complex DDIs with several statins (pitavastatin, rosuvastatin, atorvastatin, and pravastatin) by mechanistically assessing differential inhibitory effects of perpetrator drugs on multiple transporters, such as organic anion transporting polypeptides (OATP1B), breast cancer resistance protein (BCRP), multidrug resistance protein 2 (MRP2), organic anion transporter 3 (OAT3), and cytochrome P450 CYP3A enzyme, as they are known to contribute to absorption, distribution, metabolism and excretion (ADME) of above statins. These models successfully predicted a total of 46 statin DDIs, including above DAA drugs and their fix-dose combination regimens. Predicted plasma area under curve ratio (AUCR) with and without perpetrator drugs was within ~ 2-fold of observed values. In contrast, simplified static R-value model resulted in increased false negative and false positive predictions when different prediction cut-off values were applied. Our studies suggest that mechanistic static model is a promising and useful tool to provide more accurate prediction of the risk and magnitude of DDIs with statins in early drug development and may help to improve the management of clinical DDIs for HCV drugs to ensure effective and safe HCV therapy. GRAPHICAL ABSTRACT.

Enhancing drug-drug interaction prediction by three-way decision and knowledge graph embedding.

Drug–Drug interaction (DDI) prediction is essential in pharmaceutical research and clinical application. Existing computational methods mainly extract data from multiple resources and treat it as binary classification. However, this cannot unambiguously tell the boundary between positive and negative samples owing to the incompleteness and uncertainty of derived data. A granular computing method called three-way decision is proved to be effective in making uncertain decision, but it relies on supplementary information to make delay decision. Recently, biomedical knowledge graph has been regarded as an important source to obtain abundant supplementary information about drugs. This paper proposes a three-way decision-based method called 3WDDI, in combination with knowledge graph embedding as supplementary features to enhance DDI prediction. The drug pairs are divided into positive, negative and boundary regions by Convolutional Neural Network (CNN) according to drug chemical structure feature. Further, delay decision is made for objects in the boundary region by integrating knowledge graph embedding feature to promote the accuracy of decision-making. The empirical results show that 3WDDI yields up to 0.8922, 0.9614, 0.9582, 0.8930 for Accuracy, AUPR, AUC and F1-score, respectively, and outperforms several baseline models.

DeSIDE-DDI: interpretable prediction of drug-drug interactions using drug-induced gene expressions

Adverse drug-drug interaction (DDI) is a major concern to polypharmacy due to its unexpected adverse side effects and must be identified at an early stage of drug discovery and development. Many computational methods have been proposed for this purpose, but most require specific types of information, or they have less concern in interpretation on underlying genes. We propose a deep learning-based framework for DDI prediction with drug-induced gene expression signatures so that the model can provide the expression level of interpretability for DDIs. The model engineers dynamic drug features using a gating mechanism that mimics the co-administration effects by imposing attention to genes. Also, each side-effect is projected into a latent space through translating embedding. As a result, the model achieved an AUC of 0.889 and an AUPR of 0.915 in unseen interaction prediction, which is competitively very accurate and outperforms other state-of-the-art methods. Furthermore, it can predict potential DDIs with new compounds not used in training. In conclusion, using drug-induced gene expression signatures followed by gating and translating embedding can increase DDI prediction accuracy while providing model interpretability. The source code is available on GitHub (https://github.com/GIST-CSBL/DeSIDE-DDI).

Experimental and Modeling Evidence Supporting the Trans-Inhibition Mechanism for Preincubation Time-Dependent, Long-Lasting Inhibition of Organic Anion Transporting Polypeptide 1B1 by Cyclosporine A.

Cyclosporine A (CsA) and rifampin are potent inhibitors of organic anion transporting polypeptide (OATP) 1B1 and are widely used to assess the risk for drug-drug interactions. CsA displays preincubation time-dependent, long-lasting inhibition of OATP1B1 in vitro and in rats in vivo, and a proposed mechanism is the trans-inhibition by which CsA inhibits OATP1B1 from the inside of cells. The current study aimed to experimentally validate the proposed mechanism using human embryonic kidney 293 cells stably expressing OATP1B1. The uptake of CsA reached a plateau following an approximate 60-minute incubation, with the cell-to-buffer concentration ratio of 3930, reflective of the high-affinity, high-capacity intracellular binding of CsA. The time course of CsA uptake was analyzed to estimate the kinetic parameters for permeability clearance and intracellular binding. When the OATP1B1-mediated uptake of [3H]estradiol-17β-glucuronide was measured following preincubation with CsA for 5 to 120 minutes, apparent Ki values became lower with longer preincubation. Our kinetic modeling incorporated the two reversible inhibition constants [Ki,trans and Ki,cis for the inhibition from inside (trans-inhibition) and outside (cis-inhibition) of cells, respectively] and estimated Ki,trans value of CsA was smaller by 48-fold than the estimated Ki,cis value. Rifampin also displayed preincubation time-dependent inhibition of OATP1B1, albeit the extent of enhancement was only twofold. The current study provides experimental evidence for the preincubation time-dependent shift of apparent Ki values and a mechanistic basis for physiologically based pharmacokinetic modeling that incorporates permeability clearance, extensive intracellular binding, and asymmetry of Ki values between the inside and outside of cells. SIGNIFICANCE STATEMENT: In vitro data and kinetic modeling support that preincubation time-dependent, long-lasting inhibition of OATP1B1 by CsA can be explained by the extensive intracellular binding and reversible OATP1B1 inhibition intracellularly (trans-inhibition) as well as extracellularly (cis-inhibition). For inhibitors to display time-dependency, the following factors were found important: time to reach a steady-state cellular concentration, trans-inhibition potency relative to cis-inhibition, and the degree of cellular inhibitor accumulation. This study would aid in the accurate prediction of drug-drug interactions mediated by OATP1B1 inhibition.

DeepMDDI: A deep graph convolutional network framework for multi-label prediction of drug-drug interactions

It is crucial to identify DDIs and explore their underlying mechanism (e.g., DDIs types) for polypharmacy safety. However, the detection of DDIs in assays is still time-consuming and costly, due to the need for experimental search over a large space of drug combinations. Thus, many computational methods have been developed to predict DDIs, most of them focusing on whether a drug interacts with another or not. And a few deep learning-based methods address a more realistic screening task for identifying various DDI types, but they assume a DDI only triggers one pharmacological effect, while a DDI can trigger more types of pharmacological effects. Thus, here we proposed a novel end-to-end deep learning-based method (called deepMDDI) for the Multi-label prediction of Drug-Drug Interactions. deepMDDI contains an encoder derived from relational graph convolutional networks and a tensor-like decoder to uniformly model interactions. deepMDDI is not only efficient for DDI transductive prediction, but also inductive prediction. The experimental results show that our model is superior to other state-of-the-art deep learning-based methods. We also validated the power of deepMDDI in the DDIs multi-label prediction and found several new valid DDIs in the case study. In conclusion, deepMDDI is beneficial to uncover the mechanism and regularity of DDIs.

SmileGNN: Drug-Drug Interaction Prediction Based on the SMILES and Graph Neural Network.

Concurrent use of multiple drugs can lead to unexpected adverse drug reactions. The interaction between drugs can be confirmed by routine in vitro and clinical trials. However, it is difficult to test the drug–drug interactions widely and effectively before the drugs enter the market. Therefore, the prediction of drug–drug interactions has become one of the research priorities in the biomedical field. In recent years, researchers have been using deep learning to predict drug–drug interactions by exploiting drug structural features and graph theory, and have achieved a series of achievements. A drug–drug interaction prediction model SmileGNN is proposed in this paper, which can be characterized by aggregating the structural features of drugs constructed by SMILES data and the topological features of drugs in knowledge graphs obtained by graph neural networks. The experimental results show that the model proposed in this paper combines a variety of data sources and has a better prediction performance compared with existing prediction models of drug–drug interactions. Five out of the top ten predicted new drug–drug interactions are verified from the latest database, which proves the credibility of SmileGNN.

Increasing application of pediatric physiologically based pharmacokinetic models across academic and industry organizations.

There has been a significant increase in the use of physiologically based pharmacokinetic (PBPK) models during the past 20 years, especially for pediatrics. The aim of this study was to give a detailed overview of the growth and areas of application of pediatric PBPK (P‐PBPK) models. A total of 181 publications and publicly available regulatory reviews were identified and categorized according to year, author affiliation, platform, and primary application of the P‐PBPK model (in clinical settings, drug development or to advance pediatric model development in general). Secondary application areas, including dose selection, biologics, and drug interactions, were also assessed. The growth rate for P‐PBPK modeling increased 33‐fold between 2005 and 2020; this was mainly attributed to growth in clinical and drug development applications. For primary applications, 50% of articles were classified under clinical, 18% under drug development, and 33% under model development. The most common secondary applications were dose selection (75% drug development), pharmacokinetic prediction and covariate identification (47% clinical), and model parameter identification (68% model development), respectively. Although population PK modeling remains the mainstay of approaches supporting pediatric drug development, the data presented here demonstrate the widespread application of P‐PBPK models in both drug development and clinical settings. Although applications for pharmacokinetic and drug–drug interaction predictions in pediatrics is advocated, this approach remains underused in areas such as assessment of pediatric formulations, toxicology, and trial design. The increasing number of publications supporting the development and refinement of the pediatric model parameters can only serve to enhance optimal use of P‐PBPK models.

Computational investigation of Schiff bases from tryptamine as COX-2 inhibitors with potential anti-inflammatory activity

Non-steroidal anti-inflammatory drugs are the most used drugs to inhibit cyclooxygenases (COX). Understanding the structure of COX isoforms is required for the research for more selective and less toxic anti-inflammatory agents. Also, computational methods can help to predict drug interaction. In this work, tryptamine Schiff bases, retrieved from literature, were screened to evaluate their ADMET profiles. After the initial screening, the compounds with better parameters were subjected to molecular docking interactions with the crystal structure of the COX-2 enzyme to verify binding affinities and the compound’s disposition as anti-inflammatory drugs. All compounds obeyed the Veber rule, and only one of them violated the Lipinski rule. Additionally, ten tryptamine Schiff bases presented equal or fewer toxicity alerts than Celecoxib (≤1). The binding affinities obtained through molecular docking for these Schiff bases varied from -9.974 to -8.765 kcal/mol. The results demonstrated that the chosen compounds showed higher binding affinities when compared to Celecoxib (-8.655 kcal/mol); this suggests a significant ability to inhibit the COX-2 enzyme. It has been proposed that the presence of phenolic hydroxyl is significant for the observed binding affinities. The results of this study indicate that tryptamine Schiff bases are promising candidates to treat inflammatory disorders, with the potential to be used as suitable medicines for pain and inflammation treatment.

A Review of Approaches for Predicting Drug-Drug Interactions Based on Machine Learning.

Drug–drug interactions play a vital role in drug research. However, they may also cause adverse reactions in patients, with serious consequences. Manual detection of drug–drug interactions is time-consuming and expensive, so it is urgent to use computer methods to solve the problem. There are two ways for computers to identify drug interactions: one is to identify known drug interactions, and the other is to predict unknown drug interactions. In this paper, we review the research progress of machine learning in predicting unknown drug interactions. Among these methods, the literature-based method is special because it combines the extraction method of DDI and the prediction method of DDI. We first introduce the common databases, then briefly describe each method, and summarize the advantages and disadvantages of some prediction models. Finally, we discuss the challenges and prospects of machine learning methods in predicting drug interactions. This review aims to provide useful guidance for interested researchers to further promote bioinformatics algorithms to predict DDI.

Machine learning-driven identification of drugs inhibiting cytochrome P450 2C9.

Cytochrome P450 2C9 (CYP2C9) is a major drug-metabolizing enzyme that represents 20% of the hepatic CYPs and is responsible for the metabolism of 15% of drugs. A general concern in drug discovery is to avoid the inhibition of CYP leading to toxic drug accumulation and adverse drug–drug interactions. However, the prediction of CYP inhibition remains challenging due to its complexity. We developed an original machine learning approach for the prediction of drug-like molecules inhibiting CYP2C9. We created new predictive models by integrating CYP2C9 protein structure and dynamics knowledge, an original selection of physicochemical properties of CYP2C9 inhibitors, and machine learning modeling. We tested the machine learning models on publicly available data and demonstrated that our models successfully predicted CYP2C9 inhibitors with an accuracy, sensitivity and specificity of approximately 80%. We experimentally validated the developed approach and provided the first identification of the drugs vatalanib, piriqualone, ticagrelor and cloperidone as strong inhibitors of CYP2C9 with IC values <18 μM and sertindole, asapiprant, duvelisib and dasatinib as moderate inhibitors with IC50 values between 40 and 85 μM. Vatalanib was identified as the strongest inhibitor with an IC50 value of 0.067 μM. Metabolism assays allowed the characterization of specific metabolites of abemaciclib, cloperidone, vatalanib and tarafenacin produced by CYP2C9. The obtained results demonstrate that such a strategy could improve the prediction of drug-drug interactions in clinical practice and could be utilized to prioritize drug candidates in drug discovery pipelines.

Learning size-adaptive molecular substructures for explainable drug-drug interaction prediction by substructure-aware graph neural network.

Drug–drug interactions (DDIs) can trigger unexpected pharmacological effects on the body, and the causal mechanisms are often unknown. Graph neural networks (GNNs) have been developed to better understand DDIs. However, identifying key substructures that contribute most to the DDI prediction is a challenge for GNNs. In this study, we presented a substructure-aware graph neural network, a message passing neural network equipped with a novel substructure attention mechanism and a substructure–substructure interaction module (SSIM) for DDI prediction (SA-DDI). Specifically, the substructure attention was designed to capture size- and shape-adaptive substructures based on the chemical intuition that the sizes and shapes are often irregular for functional groups in molecules. DDIs are fundamentally caused by chemical substructure interactions. Thus, the SSIM was used to model the substructure–substructure interactions by highlighting important substructures while de-emphasizing the minor ones for DDI prediction. We evaluated our approach in two real-world datasets and compared the proposed method with the state-of-the-art DDI prediction models. The SA-DDI surpassed other approaches on the two datasets. Moreover, the visual interpretation results showed that the SA-DDI was sensitive to the structure information of drugs and was able to detect the key substructures for DDIs. These advantages demonstrated that the proposed method improved the generalization and interpretation capability of DDI prediction modeling.

Biomedical Knowledge Graph Embedding with Capsule Network for Multi-label Drug-Drug Interaction Prediction

Drug-drug interaction (DDI) plays an important role in drug development and administration. Most of existing network-based computation models regard the DDI prediction as a binary classification problem and generate negative DDI samples randomly, but the binary classification is not in line with the real problem since there are dozens of types of DDI and randomly generating negative samples may introduce false-negative samples since the non-observed facts can be either false or just missing. To address the above limitations, we propose a new framework called KG2ECapsule that explicitly models the multi-relational DDI data based on biomedical knowledge graphs in an end-to-end fashion. It first generates high-quality negative samples based on the average number of tail entities and head entities for each relation to reduce false-negative samples to some extent. KG2ECapsule then refines the representations of entities by recursively propagating the embeddings from the attention-based receptive fields of entities. Empirical results on three biomedical knowledge graphs of different scales show that KG2ECapsule outperforms the state-of-the-art methods consistently in multi-label DDI prediction task and further studies verify the efficacy of both probability-based sampling strategy and non-linear transformation for modeling multi-relational data.

A Simple and Rapid Bioluminescence-Based Functional Assay of Organic Anion Transporter 1 as a D-Luciferin Transporter.

Organic anion transporter 1 (SLC22A6/OAT1) plays a key role in renal tubular excretion of endo- and exogenous anionic substances including drugs. Since the inhibition of OAT1 function by a concomitant drug may cause pharmacokinetic drug-drug interactions (DDIs) in clinical practice, an in vitro uptake study to evaluate the inhibition potency of OAT1 is useful for the prediction and avoidance of DDIs and recommended for drug candidates in drug development. In this chapter, we describe a rapid and highly sensitive functional assay of OAT1 based on bioluminescence (BL) detection using D-luciferin as a substrate in living cells. The principle of measurement simply relies on the biochemical feature of D-luciferin to be recognized as a substrate of OAT1, and the BL intensity depending on intracellular D-luciferin level and luciferase activity, thereby allowing the quantitative analysis of OAT1-mediated D-luciferin transport. The BL measurement can be completed within 1min without experimental procedures for removing extracellular uptake solution and washing cells, both of which involve in the conventional uptake studies using isotope-labeled or fluorescent compounds. The present method is applicable to high-throughput screening to identify and avoid potential OAT1 inhibitors in drug development.

Application of Solubility and Dissolution Profile Comparison for Prediction of Gastric pH-Mediated Drug-Drug Interactions.

The objective of this study was to assess how solubility and dissolution profile comparisons under different pH conditions can be used to predict gastric pH-mediated drug-drug interaction (DDI) potential. We collected information for new molecular entities (NMEs) approved from 2003 to 2019 by the U.S. Food and Drug Administration (FDA) that had dedicated clinical DDI studies with acid-reducing agents (ARAs). Among these, 67 NMEs with solubility under different pHs and dissolution profiles generated in pH 1.2, 4.5, and 6.8 aqueous media were included for analysis. Similarity factor (f2) was used to compare dissolution profiles at different pHs for pH-mediated DDI prediction (e.g., f2<50 predicts positive DDI). Prediction accuracy was calculated based on the outcome comparison between predicted and observed DDIs. Based on dissolution profile comparisons and observed DDI data, weak base drugs (WBDs) (n = 49) showed 72.5% prediction accuracy under the fasted conditions, and 66.7% prediction accuracy under fed conditions. While using solubility and clinical dose for prediction, the prediction accuracy was 80% under fasted conditions and 66.7% under fed conditions, respectively. Comparison of dissolution profiles generated at pH 1.2, 4.5, and 6.8 can be used to predict gastric pH-mediated DDI potential for WBDs. It demonstrated comparable prediction accuracy under both fasted and fed conditions when compared to the prediction using solubility and clinical dose. Furthermore, dissolution profile comparison could add an additional understanding of possible impact of pH change on the release behavior of the drug product. Graphical abstract.