Prediction Of Drug-drug Interactions Research Articles

SRR-DDI: A drug–drug interaction prediction model with substructure refined representation learning based on self-attention mechanism

Drug–drug interaction (DDI) is an important safety issue during clinical treatment, where the mechanism of action of drugs may interfere with each other thereby causing adverse effects on the body leading to therapeutic failure. Since the deep learning method provide a powerful tool in DDI prediction, in this paper, we propose a DDI prediction model based on substructure Refined Representation Learning based on Self-Attention Mechanism, SRR-DDI, to improve the robustness of the substructure features determining the properties of the drug, thus improving the performance of DDI prediction. To improve the generalization of the model, the drug similarity feature is designed and introduced to help the model extract potential associations between drugs. The comparative experiments are set up on real-word data with two scenarios of warm start and cold start for performance evaluation, and the experimental results show that SRR-DDI outperforms the state-of-the-art methods. Finally, the visual interpretation experiment demonstrates that SRR-DDI can gradually refine the substructure features, and highlight the important substructures of drugs in DDI. In summary, SRR-DDI provides a powerful tool for predicting DDI and in-depth understanding of DDI.

A Novel Deep Learning Model for Drug-Drug Interactions

Drug-drug interactions (DDIs) can lead to adverse events and compromised treatment efficacy, emphasizing the need for accurate prediction and understanding of these interactions. This paper presents an innovative approach for DDI prediction, employing two distinct message-passing neural network (MPNN) models, each meticulously tailored to focus on one drug within a pair. By adeptly capturing the unique attributes and interactions of each drug, this novel method strives to enhance the accuracy of DDI prediction. Furthermore, we integrate the outcomes of individual MPNN models with information derived from both drugs and their molecular features. In addition to this, we harnessed a curated list of 82 existing FDA-approved drugs, which were evaluated against five vital SARS-CoV-2 proteins. Based on the Combined Score, we systematically scrutinized a selection of the top-10 drugs exhibiting the highest binding affinity to these proteins, employing a sophisticated deep learning architecture for predicting Drug-Drug Interactions. The results remarkably demonstrated the effectiveness of our approach, boasting an impressive accuracy of 0.92, an area under the curve (AUC) of 0.99, and an F1-score of 0.85. Moreover, the identified list of high-affinity drugs against SARS-CoV-2 proteins may potentially serve as a pivotal starting point for the development of new and invaluable pharmaceutical interventions.

Emerging drug interaction prediction enabled by a flow-based graph neural network with biomedical network.

Drug-drug interactions (DDIs) for emerging drugs offer possibilities for treating and alleviating diseases, and accurately predicting these with computational methods can improve patient care and contribute to efficient drug development. However, many existing computational methods require large amounts of known DDI information, which is scarce for emerging drugs. Here we propose EmerGNN, a graph neural network that can effectively predict interactions for emerging drugs by leveraging the rich information in biomedical networks. EmerGNN learns pairwise representations of drugs by extracting the paths between drug pairs, propagating information from one drug to the other, and incorporating the relevant biomedical concepts on the paths. The edges of the biomedical network are weighted to indicate the relevance for the target DDI prediction. Overall, EmerGNN has higher accuracy than existing approaches in predicting interactions for emerging drugs and can identify the most relevant information on the biomedical network.

GADNN: A graph attention-based method for drug-drug association prediction considering the contribution rate of different types of drug-related features

The simultaneous use of multiple drugs, known as drug combinations, is increasingly common in the treatment of complex diseases such as cancer. However, drug-drug interactions (DDIs) caused by the simultaneous use of multiple drugs can lead to unexpected side effects and even death, resulting in high medical costs. Therefore, understanding potential DDI is an essential step in reducing the risk of adverse drug reactions before administering clinical drug compounds. Recently, many machine learning-based methods for DDI prediction have been introduced, but in these methods, no attention is paid to determining the contribution rate of each type of drug feature. Here, a Graph Attention-based Deep Neural Network method called GADNN is proposed to predict DDIs based on considering the influence or contribution rate of different drug-related features. The introduced method consists of two stages. In the first stage, four basic drug datasets, including target, enzyme, infrastructure, and pathway, are used as types of drug features. Using a graph neural network-based method, the embedding vector is generated for each drug based on each dataset separately. In the second stage, the contribution coefficient of each dataset is calculated dynamically through a new graph attention mechanism. Then, the weighted combination of the separate vectors is used to predict the probability of drug-drug interactions using a dense neural network. The experimental results confirmed that the presented innovation has been able to improve the accuracy compared to state-of-the-art methods.

Assessment of CYP3A-mediated drug interaction via cytokine (IL-6) elevation for mosunetuzumab using physiologically-based pharmacokinetic modeling.

Mosunetuzumab is a CD3/CD20 bispecific antibody. As an on-target effect, transient elevation of interleukin-6 (IL-6) occurs in early treatment cycles. A physiologically-based pharmacokinetic (PBPK) model was developed to assess potential drug interaction caused by IL-6 enzyme suppression on cytochrome P450 3A (CYP3A) during mosunetuzumab treatment. The model's performance in predicting IL-6 CYP3A suppression and subsequent drug-drug interactions (DDIs) was verified using existing clinical data of DDIs caused by chronic and transient IL-6 elevation. Sensitivity analyses were performed for a complete DDI risk assessment. The IL-6 concentration- and time-dependent CYP3A suppression during mosunetuzumab treatment was simulated using PBPK model with incorporation of invitro IL-6 inhibition data. At clinically approved doses/regimens, the DDI at maximum CYP3A suppression was predicted to be a midazolam maximum drug concentration in plasma (Cmax ) and area under the plasma drug concentration-time curve (AUC) ratio of 1.17 and 1.37, respectively. At the 95th percentile of IL-6 concentration level or when gut CYP3A suppression was considered, the predicted DDI risk for mosunetuzumab remained low (<2-fold). The PBPK-based DDI predictions informed the mosunetuzumab product label to monitor, in early cycles, the concentrations and toxicities for sensitive CYP3A substrates with narrow therapeutic windows.

Quantitative prediction of CYP3A-mediated drug-drug interactions by correctly estimating fraction metabolized using human liver chimeric mice.

Fraction metabolized (fm ) and fraction transported (ft ) are important for understanding drug-drug interactions (DDIs) in drug discovery and development. However, current in vitro systems cannot accurately estimate in vivo fm due to inability to reflect the ft by efflux transporters (ft,efflux ). This study demonstrates how CYP3A-mediated DDI for CYP3A/P-gp substrates can be predicted using Hu-PXB mice as human liver chimeric mice. For estimating human in vitro fm by CYP3A enzyme (fm,CYP3A,in vitro ), six drugs, including CYP3A/P-gp substrates (alprazolam, cyclosporine, docetaxel, midazolam, prednisolone, and theophylline) and human hepatocytes were incubated with or without ketoconazole as a CYP3A inhibitor. We calculated fm,CYP3A,in vitro based on hepatic intrinsic clearance. To estimate human in vivo fm,CYP3A (fm,CYP3A,in vivo ), we collected information on clinical DDI caused by ketoconazole for these six drugs. We calculated fm,CYP3A,in vivo using the change of total clearance (CLtotal ). For evaluating the human DDI predictability, the six drugs were administered intravenously to Hu-PXB and SCID mice with or without ketoconazole. We calculated the change of CLtotal caused by ketoconazole. We compared the CLtotal change in humans with that in Hu-PXB and SCID mice. The fm,CYP3A,in vitro was overestimated compared to the fm,CYP3A,in vivo . Hu-PXB mice showed much better correlation in the change of CLtotal with humans (R2 = 0.95) compared to SCID mice (R2 = 0.0058). CYP3A-mediated DDI can be predicted by correctly estimating human fm,CYP3A,in vivo using Hu-PXB mice. These mice could be useful predicting hepatic fm and ft,efflux .

GGI-DDI: Identification for key molecular substructures by granule learning to interpret predicted drug–drug interactions

Deep learning-based approaches have achieved promising performance in predicting drug–drug interactions (DDIs). Nevertheless, a significant drawback of these approaches is their limited interpretability, hindering their practical applicability for clinicians. Based on current research findings, drug interactions frequently arise from specific substructures or functional groups present in drugs. To enhance the interpretability of deep learning models, we propose a novel end-to-end learning approach that employs granular computing to identify pivotal substructures instead of using conventional atom-based or predefined molecular fingerprint methods to predict DDIs. We refer to this model as ”GGI-DDI” (Granule–Granule Interaction for Drug–Drug Interaction). In this method, drugs are granulated into a set of coarser granules that represent the key substructures or functional groups of drugs. Subsequently, the detection of DDIs occurs through the examination of interactions among these granules, aligning more closely with human cognitive patterns. We conducted thorough experiments on the TWOSIDES dataset, and the results show that GGI-DDI achieved impeccable accuracy in predicting DDIs. We compared GGI-DDI to state-of-the-art baseline models including DDIMDL, GoGNN, DNN, STNN-DDI and GMPNN-CS, GGI-DDI almost consistently outperforms the baselines across all metrics in terms of Accuracy (Acc), Area under the receiver operating characteristic (Auc), Area under precision recall curve (Aupr) and Precision (Pre) in both transductive and inductive scenarios. Finally, we provide case studies to illustrate how GGI-DDI can effectively reveal important substructure pairs across drugs about a specific DDI type, offering insights into the underlying mechanism of these interactions. We find that this interpretability can serve as valuable guidance in the advancement of novel drug development and poly-drug therapy strategies.

Enhancing drug discovery in schizophrenia: a deep learning approach for accurate drug-target interaction prediction – DrugSchizoNet

Drug discovery relies on the precise prognosis of drug–target interactions (DTI). Due to their ability to learn from raw data, deep learning (DL) methods have displayed outstanding performance over traditional approaches. However, challenges such as imbalanced data, noise, poor generalization, high cost, and time-consuming processes hinder progress in this field. To overcome the above challenges, we propose a DL-based model termed DrugSchizoNet for drug interaction (DI) prediction of Schizophrenia. Our model leverages drug-related data from the DrugBank and repoDB databases, employing three key preprocessing techniques. First, data cleaning eliminates duplicate or incomplete entries to ensure data integrity. Next, normalization is performed to enhance security and reduce costs associated with data acquisition. Finally, feature extraction is applied to improve the quality of input data. The three layers of the DrugSchizoNet model are the input, hidden and output layers. In the hidden layer, we employ dropout regularization to mitigate overfitting and improve generalization. The fully connected (FC) layer extracts relevant features, while the LSTM layer captures the sequential nature of DIs. In the output layer, our model provides confidence scores for potential DIs. To optimize the prediction accuracy, we utilize hyperparameter tuning through OB-MOA optimization. Experimental results demonstrate that DrugSchizoNet achieves a superior accuracy of 98.70%. The existing models, including CNN-RNN, DANN, CKA-MKL, DGAN, and CNN, across various evaluation metrics such as accuracy, recall, specificity, precision, F1 score, AUPR, and AUROC are compared with the proposed model. By effectively addressing the challenges of imbalanced data, noise, poor generalization, high cost and time-consuming processes, DrugSchizoNet offers a promising approach for accurate DTI prediction in Schizophrenia. Its superior performance demonstrates the potential of DL in advancing drug discovery and development processes.

A simplified similarity-based approach for drug-drug interaction prediction.

Drug-drug interactions (DDIs) are a critical component of drug safety surveillance. Laboratory studies aimed at detecting DDIs are typically difficult, expensive, and time-consuming; therefore, developing in-silico methods is critical. Machine learning-based approaches for DDI prediction have been developed; however, in many cases, their ability to achieve high accuracy relies on data only available towards the end of the molecule lifecycle. Here, we propose a simple yet effective similarity-based method for preclinical DDI prediction where only the chemical structure is available. We test the model on new, unseen drugs. To focus on the preclinical problem setting, we conducted a retrospective analysis and tested the models on drugs that were added to a later version of the DrugBank database. We extend an existing method, adjacency matrix factorization with propagation (AMFP), to support unseen molecules by applying a new lookup mechanism to the drugs' chemical structure, lookup adjacency matrix factorization with propagation (LAMFP). We show that using an ensemble of different similarity measures improves the results. We also demonstrate that Chemprop, a message-passing neural network, can be used for DDI prediction. In computational experiments, LAMFP results in high accuracy, with an area under the receiver operating characteristic curve of 0.82 for interactions involving a new drug and an existing drug and for interactions involving only existing drugs. Moreover, LAMFP outperforms state-of-the-art, complex graph neural network DDI prediction methods.

Predicting drug-drug interactions based on multi-view and multichannel attention deep learning.

Predicting drug-drug interactions (DDIs) has become a major concern in the drug research field because it helps explore the pharmacological function of drugs and enables the development of new therapeutic drugs. Existing prediction methods simply integrate multiple drug attributes or perform tasks on a biomedical knowledge graph (KG). Though effective, few methods can fully utilize multi-source drug data information. In this paper, a multi-view and multichannel attention deep learning (MMADL) model is proposed, which not only extracts rich drug features containing both drug attributes and drug-related entity information from multi-source databases, but also considers the consistency and complementarity of different drug feature representation learning approaches to improve the effectiveness and accuracy of DDI prediction. A single-layer perceptron encoder is applied to encode multi-source drug information to obtain multi-view drug representation vectors in the same linear space. Then, the multichannel attention mechanism is introduced to obtain the attention weight by adaptively learning the importance of drug features according to their contributions to DDI prediction. Further, the representation vectors of multi-view drug pairs with attention weights are used as inputs of the deep neural network to predict potential DDI. The accuracy and precision-recall curves of MMADL are 93.05 and 95.94, respectively. The results indicate that the proposed method outperforms other state-of-the-art methods.

Development of a Novel In Vitro Model to Study Lymphatic Uptake of Drugs via Artificial Chylomicrons.

The lymphatic system plays a crucial role in the absorption of lipophilic drugs, making it an important route for drug delivery. In this study, an in vitro model using Intralipid® was developed to investigate the lymphatic uptake of drugs. The model was validated using cannabidiol, halofantrine, quercetin, and rifampicin. Remarkably, the uptake of these drugs closely mirrored what would transpire in vivo. Furthermore, adding peanut oil to the model system significantly increased the lymphatic uptake of rifampicin, consistent with meals containing fat stimulating lymphatic drug uptake. Conversely, the inclusion of pluronic L-81 was observed to inhibit the lymphatic uptake of rifampicin in the model. This in vitro model emerges as a valuable tool for investigating and predicting drug uptake via the lymphatic system. It marks the first phase in developing a physiologically based predictive tool that can be refined further to enhance the precision of drug interaction predictions with chylomicrons and their subsequent transport via the lymphatic system. Moreover, it can be employed to explore innovative drug formulations and excipients that either enhance or hinder lymphatic drug uptake. The insights gained from this study have significant implications for advancing drug delivery through the lymphatic system.

Effect of probenecid on blood levels and renal elimination of furosemide and endogenous compounds in rats: Discovery of putative organic anion transporter biomarkers

Transporter-mediated drug-drug interactions (DDIs) are assessed using probe drugs and in vitro and in vivo models during drug development. The utility of endogenous metabolites as transporter biomarkers is emerging for prediction of DDIs during early phases of clinical trials. Endogenous metabolites such as pyridoxic acid and kynurenic acid have shown potential to predict DDIs mediated by organic anion transporters (OAT1 and OAT3). However, these metabolites have not been assessed in rats as potential transporter biomarkers. We carried out a rat pharmacokinetic DDI study using probenecid and furosemide as OAT inhibitor and substrate, respectively. Probenecid administration led to a 3.8-fold increase in the blood concentrations and a 3-fold decrease in renal clearance of furosemide. High inter-individual and intra-day variability in pyridoxic acid and kynurenic acid, and no or moderate effect of probenecid administration on these metabolites suggest their limited utility for prediction of Oat-mediated DDI in rats. Therefore, rat blood and urine samples were further analysed using untargeted metabolomics. Twenty-one m/z features (out of >8000 detected features) were identified as putative biomarkers of rat Oat1 and Oat3 using a robust biomarker qualification approach. These m/z features belong to metabolic pathways such as fatty acid analogues, peptides, prostaglandin analogues, bile acid derivatives, flavonoids, phytoconstituents, and steroids, and can be used as a panel to decrease variability caused by processes other than Oats. When validated, these putative biomarkers will be useful in predicting DDIs caused by Oats in rats.

MATT-DDI: Predicting multi-type drug-drug interactions via heterogeneous attention mechanisms

The joint use of multiple drugs can result in adverse drug-drug interactions (DDIs) and side effects that harm the body. Accurate identification of DDIs is crucial for avoiding accidental drug side effects and understanding potential mechanisms underlying DDIs. Several computational methods have been proposed for multi-type DDI prediction, but most rely on the similarity profiles of drugs as the drug feature vectors, which may result in information leakage and overoptimistic performance when predicting interactions between new drugs. To address this issue, we propose a novel method, MATT-DDI, for predicting multi-type DDIs based on the original feature vectors of drugs and multiple attention mechanisms. MATT-DDI consists of three main modules: the top k most similar drug pair selection module, heterogeneous attention mechanism module and multi‑type DDI prediction module. Firstly, based on the feature vector of the input drug pair (IDP), k drug pairs that are most similar to the input drug pair from the training dataset are selected according to cosine similarity between drug pairs. Then, the vectors of k selected drug pairs are averaged to obtain a new drug pair (NDP). Next, IDP and NDP are fed into heterogeneous attention modules, including scaled dot product attention and bilinear attention, to extract latent feature vectors. Finally, these latent feature vectors are taken as input of the classification module to predict DDI types. We evaluated MATT-DDI on three different tasks. The experimental results show that MATT-DDI provides better or comparable performance compared to several state-of-the-art methods, and its feasibility is supported by case studies. MATT-DDI is a robust model for predicting multi-type DDIs with excellent performance and no information leakage.

Applied physiologically-based pharmacokinetic modeling to assess uridine diphosphate-glucuronosyltransferase-mediated drug-drug interactions for Vericiguat.

Vericiguat (Verquvo; US: Merck, other countries: Bayer) is a novel drug for the treatment of chronic heart failure. Preclinical studies have demonstrated that the primary route of metabolism for vericiguat is glucuronidation, mainly catalyzed by uridine diphosphate-glucuronosyltransferase (UGT)1A9 and to a lesser extent UGT1A1. Whereas a drug-drug interaction (DDI) study of the UGT1A9 inhibitor mefenamic acid showed a 20% exposure increase, the effect of UGT1A1 inhibitors has not been assessed clinically. This modeling study describes a physiologically-based pharmacokinetic (PBPK) approach to complement the clinical DDI liability assessment and support prescription labeling. A PBPK model of vericiguat was developed based on invitro and clinical data, verified against data from the mefenamic acid DDI study, and applied to assess the UGT1A1 DDI liability by running an in silico DDI study with the UGT1A1 inhibitor atazanavir. A minor effect with an area under the plasma concentration-time curve (AUC) ratio of 1.12 and a peak plasma concentration ratio of 1.04 was predicted, which indicates that there is no clinically relevant DDI interaction anticipated. Additionally, the effect of potential genetic polymorphisms of UGT1A1 and UGT1A9 was evaluated, which showed that an average modest increase of up to 1.7-fold in AUC may be expected in the case of concomitantly reduced UGT1A1 and UGT1A9 activity for subpopulations expressing non-wild-type variants for both isoforms. This study is a first cornerstone to qualify the PK-Sim platform for use of UGT-mediated DDI predictions, including PBPK models of perpetrators, such as mefenamic acid and atazanavir, and sensitive UGT substrates, such as dapagliflozin and raltegravir.

Mechanistic Static Model based Prediction of Transporter Substrate Drug-Drug Interactions Utilizing Atorvastatin and Rifampicin.

An in vitro relative activity factor (RAF) technique combined with mechanistic static modeling was examined to predict drug-drug interaction (DDI) magnitude and analyze contributions of different clearance pathways in complex DDIs involving transporter substrates. Atorvastatin and rifampicin were used as a model substrate and inhibitor pair. In vitro studies were conducted with transfected HEK293 cells, hepatocytes and human liver microsomes. Prediction success was defined as predictions being within twofold of observations. The RAF method successfully translated atorvastatin uptake from transfected cells to hepatocytes, demonstrating its ability to quantify transporter contributions to uptake. Successful translation of atorvastatin's in vivo intrinsic hepatic clearance (CLint,h,in vivo) from hepatocytes to liver was only achieved through consideration of albumin facilitated uptake or through application of empirical scaling factors to transporter-mediated clearances. Transporter protein expression differences between hepatocytes and liver did not affect CLint,h,in vivo predictions. By integrating cis and trans inhibition of OATP1B1/OATP1B3, atorvastatin-rifampicin (single dose) DDI magnitude could be accurately predicted (predictions within 0.77-1.0 fold of observations). Simulations indicated that concurrent inhibition of both OATP1B1 and OATP1B3 caused approximately 80% of atorvastatin exposure increases (AUCR) in the presence of rifampicin. Inhibiting biliary elimination, hepatic metabolism, OATP2B1, NTCP, and basolateral efflux are predicted to have minimal to no effect on AUCR. This study demonstrates the effective application of a RAF-based translation method combined with mechanistic static modeling for transporter substrate DDI predictions and subsequent mechanistic interpretation.

Utilization of OATP1B Biomarker Coproporphyrin-I to Guide Drug-Drug Interaction Risk Assessment: Evaluation by the Pharmaceutical Industry.

Drug-drug interactions (DDIs) involving hepatic organic anion transporting polypeptides 1B1/1B3 (OATP1B) can be substantial, however, challenges remain for predicting interaction risk. Emerging evidence suggests that endogenous biomarkers, particularly coproporphyrin-I (CP-I), can be used to assess in vivo OATP1B activity. The present work under the International Consortium for Innovation and Quality in Pharmaceutical Development was aimed primarily at assessing CP-I as a biomarker for informing OATP1B DDI risk. Literature and unpublished CP-I data along with pertinent in vitro and clinical DDI information were collected to identify DDIs primarily involving OATP1B inhibition and assess the relationship between OATP1B substrate drug and CP-I exposure changes. Static models to predict changes in exposure of CP-I, as a selective OATP1B substrate, were also evaluated. Significant correlations were observed between CP-I area under the curve ratio (AUCR) or maximum concentration ratio (Cmax R) and AUCR of substrate drugs. In general, the CP-I Cmax R was equal to or greater than the CP-I AUCR. CP-I Cmax R < 1.25 was associated with absence of OATP1B-mediated DDIs (AUCR < 1.25) with no false negative predictions. CP-I Cmax R < 2 was associated with weak OATP1B-mediated DDIs (AUCR < 2). A correlation was identified between CP-I exposure changes and OATP1B1 static DDI predictions. Recommendations for collecting and interpreting CP-I data are discussed, including a decision tree for guiding DDI risk assessment. In conclusion, measurement of CP-I is recommended to inform OATP1B inhibition potential. The current analysis identified changes in CP-I exposure that may be used to prioritize, delay, or replace clinical DDI studies.

Quantitative prediction of transporter-mediated drug-drug interactions using the mechanistic static pharmacokinetic (MSPK) model

Guidance/guidelines on drug-drug interactions (DDIs) have been issued in Japan, the United States, and Europe. These guidance/guidelines provide decision trees for conducting metabolizing enzyme-mediated clinical DDI studies; however, the decision trees for transporter-mediated DDIs lack quantitative prediction methods. In this study, the accuracy of a net-effect mechanistic static pharmacokinetics (MSPK) model containing the fraction transported (ft) of transporters was examined to predict transporter-mediated DDIs. This study collected information on 25 oral drugs with new active reagents that were used in clinical DDI studies as perpetrators (42 cases) from drugs approved in Japan between April 2016 and June 2020. The AUCRs (AUC ratios with and without perpetrators) of victim drugs were predicted using the net-effect MSPK model. As a result, 83 and 95% of the predicted AUCRs were within 1.5- and 2-fold error in the observed AUCRs, respectively. In cases where the victims were statins in which pharmacokinetics several transporters are involved, 70 and 91% of the predicted AUCRs were within 1.5- and 2-fold errors, respectively. Therefore, the net-effect MSPK model was applicable for predicting the AUCRs of victims, which are substrates for multiple transporters.

Enhancing drug-drug Interaction Prediction by Integrating Physiologically-Based Pharmacokinetic Model with Fraction Metabolized by CYP3A4

ABSTRACT Background Enhancing the precision of drug–drug interaction (DDI) prediction is essential for improving drug safety and efficacy. The aim is to identify the most effective fraction metabolized by CY3A4 (f m ) for improving DDI prediction using physiologically based pharmacokinetic (PBPK) models. Research Design and Methods The f m values were determined for 33 approved drugs using a human liver microsome for in vitro measurements and the ADMET Predictor software for in silico predictions. Subsequently, these f m values were integrated into PBPK models using the GastroPlus platform. The PBPK models, combined with a ketoconazole model, were utilized to predict AUCR (AUCcombo with ketoconazole/AUCdosing alone), and the accuracy of these predictions was evaluated by comparison with observed AUCR. Results The integration of in vitro f m method demonstrates superior performance compared to the in silico f m method and f m of 100% method. Under the Guest-limits criteria, the integration of in vitro f m achieves an accuracy of 76%, while the in silico f m and f m of 100% methods achieve accuracies of 67% and 58%, respectively. Conclusions Our study highlights the importance of in vitro f m data to improve the accuracy of predicting DDIs and demonstrates the promising potential of in silico f m in predicting DDIs.

DDI-SSL: Drug–Drug Interaction Prediction Based on Substructure Signature Learning

Drug–drug interactions (DDIs) are entities composed of different chemical substructures (functional groups). In existing methods that predict drug–drug interactions based on the usage of substructures, each node is perceived as the epicenter of a sub-pattern, and adjacent nodes eventually become centers of similar substructures, resulting in redundancy. Furthermore, the significant differences in structure and properties among compounds can lead to unrelated pairings, making it difficult to integrate information. This heterogeneity negatively affects the prediction results. In response to these challenges, we propose a drug–drug interaction prediction method based on substructure signature learning (DDI-SSL). This method extracts useful information from local subgraphs surrounding drugs and effectively utilizes substructures to assist in predicting drug side effects. Additionally, a deep clustering algorithm is used to aggregate similar substructures, allowing any individual subgraph to be reconstructed using this set of global signatures. Furthermore, we developed a layer-independent collaborative attention mechanism to model the mutual influence between drugs, generating signal strength scores for each class of drugs to mitigate noise caused by heterogeneity. Finally, we evaluated DDI-SSL on a comprehensive dataset and demonstrated improved performance in DDI prediction compared to state-of-the-art methods.

MM-GANN-DDI: Multimodal Graph-Agnostic Neural Networks for Predicting Drug–Drug Interaction Events

Personalized treatment of complex diseases relies on combined medication. However, the occurrence of unexpected drug–drug interactions (DDIs) in these combinations can lead to adverse effects or even fatalities. Although recent computational methods exhibit promising performance in DDI screening, their practical implementation faces two significant challenges: (i) the availability of comprehensive datasets to support clinical application, and (ii) the ability to infer DDI types for new drugs beyond the existing dataset coverage. To mitigate these challenges, we propose MM-GANN-DDI: a Multimodal Graph-Agnostic Neural Network for Predicting Drug–Drug Interaction Events. We first mine six drug modalities and incorporate a graph attention (GAT) mechanism to fuse these modalities with the topological features of the DDI graph. We further propose a novel graph neural network training mechanism called graph-agnostic meta-training (GAMT), which effectively leverages topological information from the DDI graph and efficiently predicts DDI types for new drugs beyond the available dataset. Specifically, GAMT samples meta-graphs from the original DDI graph, splitting them into support and query sets to simulate seen and unseen drugs. Two-level optimizations are applied to enhance the model’s generalization capability.We evaluate our model on two datasets (DB-v1 and DB-v2) across three tasks. Our MM-GANN-DDI demonstrates competitive performance on all three tasks. Notably, in Task 2, which focuses on predicting DDI types for drugs outside the dataset, our proposed model outperforms other methods, exhibiting an improvement of 4.6 percentage points in AUPR on DB-v1 and 5.9 percentage points on DB-v2. Additionally, our model surpasses state-of-the-art methods and classic approaches in terms of accuracy, F1 score, precision, and recall. Ablation experiments provide further validation of the effectiveness of the proposed model design. Importantly, our model exhibits the potential to discover unobserved DDIs, demonstrating its practical application in clinical medication.