Capsule Networks Showed Excellent Performance in the Classification of hERG Blockers/Nonblockers.

Yiwei Wang,Siwen Jiang,Yifei Wang,Jun Zou,Hongguang Fu,Lei Huang,Shengyong Yang

doi:10.3389/fphar.2019.01631

Abstract

Capsule networks (CapsNets), a new class of deep neural network architectures proposed recently by Hinton et al., have shown a great performance in many fields, particularly in image recognition and natural language processing. However, CapsNets have not yet been applied to drug discovery-related studies. As the first attempt, we in this investigation adopted CapsNets to develop classification models of hERG blockers/nonblockers; drugs with hERG blockade activity are thought to have a potential risk of cardiotoxicity. Two capsule network architectures were established: convolution-capsule network (Conv-CapsNet) and restricted Boltzmann machine-capsule networks (RBM-CapsNet), in which convolution and a restricted Boltzmann machine (RBM) were used as feature extractors, respectively. Two prediction models of hERG blockers/nonblockers were then developed by Conv-CapsNet and RBM-CapsNet with the Doddareddy's training set composed of 2,389 compounds. The established models showed excellent performance in an independent test set comprising 255 compounds, with prediction accuracies of 91.8 and 92.2% for Conv-CapsNet and RBM-CapsNet models, respectively. Various comparisons were also made between our models and those developed by other machine learning methods including deep belief network (DBN), convolutional neural network (CNN), multilayer perceptron (MLP), support vector machine (SVM), k-nearest neighbors (kNN), logistic regression (LR), and LightGBM, and with different training sets. All the results showed that the models by Conv-CapsNet and RBM-CapsNet are among the best classification models. Overall, the excellent performance of capsule networks achieved in this investigation highlights their potential in drug discovery-related studies.

Highlights

The human ether-a-go-go-related gene encodes a potassium channel protein, which is important for cardiac electrical activity and the coordination of heartbeat
Architecture The architecture of Conv-CapsNet is schematically shown in Figure 1, which is similar in nature to that of the Hinton's original Capsule Network, except for one additional hidden feature layer
To find the optimal capsule network architectures for the modeling of human ether-a-go-go-related gene (hERG) blockade, we tried to construct a number of capsule networks with different architectures following Hinton's principle

Summary

Introduction

The human ether-a-go-go-related gene (hERG) encodes a potassium channel protein, which is important for cardiac electrical activity and the coordination of heartbeat. Various in vitro experimental assays, such as fluorescent measurements (Dorn et al, 2005), radioligand binding assay (Yu et al, 2014), and patch-clamp electrophysiology (Stoelzle et al, 2011; Gillie et al, 2013; Danker and Moller, 2014), have been developed to measure the hERG binding affinity of chemicals. These assays are often expensive and time-consuming, implying that they are not suitable for the evaluation of hERG binding affinity for a large number of chemicals in the early stage of drug discovery. An alternative strategy is to use in silico methods; compared with experimental assays, in silico methods are cheaper and faster, and do not involve any of the aforementioned preconditions

Methods

Results

Conclusion