MDDI-SCL: predicting multi-type drug-drug interactions via supervised contrastive learning

Abstract

The joint use of multiple drugs may cause unintended drug-drug interactions (DDIs) and result in adverse consequence to the patients. Accurate identification of DDI types can not only provide hints to avoid these accidental events, but also elaborate the underlying mechanisms by how DDIs occur. Several computational methods have been proposed for multi-type DDI prediction, but room remains for improvement in prediction performance. In this study, we propose a supervised contrastive learning based method, MDDI-SCL, implemented by three-level loss functions, to predict multi-type DDIs. MDDI-SCL is mainly composed of three modules: drug feature encoder and mean squared error loss module, drug latent feature fusion and supervised contrastive loss module, multi-type DDI prediction and classification loss module. The drug feature encoder and mean squared error loss module uses self-attention mechanism and autoencoder to learn drug-level latent features. The drug latent feature fusion and supervised contrastive loss module uses multi-scale feature fusion to learn drug pair-level latent features. The prediction and classification loss module predicts DDI types of each drug pair. We evaluate MDDI-SCL on three different tasks of two datasets. Experimental results demonstrate that MDDI-SCL achieves better or comparable performance as the state-of-the-art methods. Furthermore, the effectiveness of supervised contrastive learning is validated by ablation experiment, and the feasibility of MDDI-SCL is supported by case studies. The source codes are available at https://github.com/ShenggengLin/MDDI-SCL.

Application of Artificial Intelligence in Drug–Drug Interactions Prediction: A Review

“…However, it is important to note that drug−target interaction prediction using machine learning is a dichotomous problem and cannot accurately predict the specific type of action of a drug target. Nonetheless, this approach remains a promising avenue for accelerating the 31 Chemical structural, Side effects Similarity, Logistic regression Undirected DDI Ferdousi et al 32 Carriers, Transporters Enzymes, Targets Similarity, Conventional classifier Undirected DDI Li et al 35 Chemical structural Similarity, Bayesian network Undirected DDI Kim et al 36 Medical terms, Semantic information SVM classifier, Huber loss function, Undirected DDI Cheng et al 37 Chemical structural Phenotypic Similarity, Conventional classifier Undirected DDI Park et al 45 Drug, Protein Similarity, Random walk Undirected DDI Zhang et al 61 Drug features Similarity, Matrix factorization Undirected DDI Rohani et al 51 Drug features Similarity, Matrix factorization Undirected DDI Kastrin et al 62 Chemical structural, Enzymes, Targets, Pathway Neighborhood recommendation, Random walk Undirected DDI Yan et al 38 Chemical structural, Biological, Phenotypic Similarity, RLS classifier Undirected DDI Ryu et al 78 Chemical structural Deep Neural Network, Multitask DDI events Lee et al 79 Chemical structural, Target gene, GO terms Deep Neural Network, Autoencoders DDI events Hou et al 100 Chemical structural Deep Neural Network DDI events Huang et al 101 Chemical structural Similarity, Graph neural networks DDI events Deng et al 12 Chemical structural, Enzymes, Targets, Pathway Similarity, Deep Neural Network DDI events Wang et al 102 Chemical structural, Enzymes, Targets Transformer, Autoencoders DDI events Lyu et al 103 Chemical structural, Enzymes, Targets Knowledge graph, Graph neural networks DDI events Zhu et al 114 Chemical structural Dual-view, Graph neural networks DDI events Deng et al 104 Chemical structural Small-sample learning DDI events Kang et al 105 Chemical structural, Enzymes, Targets, Pathway Deep Neural Network, Graph neural networks DDI events Shao et al 106 Chemical structural, Semantic information Pretrained transformer DDI events Lin et al 107 Drug features Attention, Contrastive learning DDI events Feng et al 29 Chemical drug discovery process and identifying potential therapeutic targets. 109 Zhang et al proposed a method for predicting drug target interactions us...…”

“…The supervised contrastive learning method has also been used to anticipate drug interactions events. Lin et al 107 proposed the MDDI-SCL in 2022, which utilizes a three-level loss function to predict multitype drug−drug interactions through supervised contrastive learning. The model consists of three modules: a drug feature encoder and mean squared error loss module that learns the latent features of drugs using selfattention mechanisms and autoencoders, a drug latent feature fusion supervised contrastive loss module that learns the latent features of drug pairs using multiscale fusion, and a drug interaction event module for classification and prediction that determines the DDI type for each drug pair.…”

Sequential Contrastive and Deep Learning Models to Identify Selective Butyrylcholinesterase Inhibitors

“…26,27 Supervised CL uses the same learning schema but allows for labeling of the data in order to generate a more accurate representation of data points and provide better classification accuracy. 28 Variations of supervised CL have been applied to drug-disease, 29 drug-target, 30−32 and multitype drug−drug 33 interaction problems in the drug discovery field as well as many other areas of cheminformatics. 34−36 .…”

“…Contrastive learning (CL) is a machine learning algorithm that was originally developed for unsupervised machine learning tasks, to classify objects as “positive” or “negative” based on similarities and differences that exist within the data, instead of known labels place on the objects beforehand. , Supervised CL uses the same learning schema but allows for labeling of the data in order to generate a more accurate representation of data points and provide better classification accuracy . Variations of supervised CL have been applied to drug-disease, drug-target, − and multitype drug–drug interaction problems in the drug discovery field as well as many other areas of cheminformatics. − . − In the study presented here, we created a supervised CL model to predict selective BChE inhibition. We compared this model’s performance to that of a deep learning (DL) model comprised of a Long Short-term memory , (LSTM) module and multilayer perceptron (MLP) and an RF model.…”

DPSP: a multimodal deep learning framework for polypharmacy side effects prediction

“…Polypharmacy (the concurrent use of multiple drugs) has become a successful strategy for combating complex or co-existing diseases ( Masnoon et al 2017 , Guillot et al 2020 , Masumshah et al 2021 , Tanvir et al 2022 , Yao et al 2022 , Lin et al 2022a ). However, in some cases, drug combinations can cause Adverse Drug Reactions (i.e.…”