Prediction Models for Agonists and Antagonists of Molecular Initiation Events for Toxicity Pathways Using an Improved Deep-Learning-Based Quantitative Structure–Activity Relationship System

Abstract

In silico approaches have been studied intensively to assess the toxicological risk of various chemical compounds as alternatives to traditional in vivo animal tests. Among these approaches, quantitative structure–activity relationship (QSAR) analysis has the advantages that it is able to construct models to predict the biological properties of chemicals based on structural information. Previously, we reported a deep learning (DL) algorithm-based QSAR approach called DeepSnap-DL for high-performance prediction modeling of the agonist and antagonist activity of key molecules in molecular initiating events in toxicological pathways using optimized hyperparameters. In the present study, to achieve high throughput in the DeepSnap-DL system–which consists of the preparation of three-dimensional molecular structures of chemical compounds, the generation of snapshot images from the three-dimensional chemical structures, DL, and statistical calculations—we propose an improved DeepSnap-DL approach. Using this improved system, we constructed 59 prediction models for the agonist and antagonist activity of key molecules in the Tox21 10K library. The results indicate that modeling of the agonist and antagonist activity with high prediction performance and high throughput can be achieved by optimizing suitable parameters in the improved DeepSnap-DL system.

A Deep Learning-Based Quantitative Structure–Activity Relationship System Construct Prediction Model of Agonist and Antagonist with High Performance

“…In addition, the highest prediction performance values of F on the valid dataset for the angles and datasplit ratios were 0.935 at 176 • and train:valid:test = 7:1:2 in 720725_GR_ant, 0.914 at 176 • and train:valid:test = 3:1:2 in 1347030_TRHR_ago, and 0.876 at 176 • and train:valid:test = 7:1:2 in 1347032_TGF_beta_ant (Figures S11 and S12; Table 3). In this study, we observed two performance peaks in prediction models at 176 • and 355 • of angles in DeepSnap, according to previous results [53,54]. These findings suggested that image augmentation is effectively worked.…”

“…The prediction models using the DeepSnap-DL system achieved higher performance than conventional ML techniques, such as random forest, XGBoost, LightGBM, and CatBoost [51,52]. Additionally, prediction models of MIE molecule agonist or antagonist activity were constructed using by the DeepSnap-DL system with the Tox21 10k library, suggesting this system as essential tool for novel QSAR analysis due to automatic feature extraction with numerous structural information from a 3D-chemical structure [53,54]. For high-throughput of the DeepSnap-DL system, automation in the DeepSanp-DL system has been conducted by combining each process consisting of the generation of images from a 3D-chemical structure based on the simplified molecular input line entry system (SMILES) format, DL using these images as input data, and calculation of prediction-performance indexes using TensorFlow and Keras [54].…”

“…Additionally, prediction models of MIE molecule agonist or antagonist activity were constructed using by the DeepSnap-DL system with the Tox21 10k library, suggesting this system as essential tool for novel QSAR analysis due to automatic feature extraction with numerous structural information from a 3D-chemical structure [53,54]. For high-throughput of the DeepSnap-DL system, automation in the DeepSanp-DL system has been conducted by combining each process consisting of the generation of images from a 3D-chemical structure based on the simplified molecular input line entry system (SMILES) format, DL using these images as input data, and calculation of prediction-performance indexes using TensorFlow and Keras [54]. In the modified DeepSnap-DL system, the mean values of receiver operating characteristic area under the curve (ROC_AUC) of the prediction models for 59 MIE targets in validation, test, and foldout datasets indicated 0.818 ± 0.056, 0.803 ± 0.063, and 0.792 ± 0.076, respectively [54].…”

“…For high-throughput of the DeepSnap-DL system, automation in the DeepSanp-DL system has been conducted by combining each process consisting of the generation of images from a 3D-chemical structure based on the simplified molecular input line entry system (SMILES) format, DL using these images as input data, and calculation of prediction-performance indexes using TensorFlow and Keras [54]. In the modified DeepSnap-DL system, the mean values of receiver operating characteristic area under the curve (ROC_AUC) of the prediction models for 59 MIE targets in validation, test, and foldout datasets indicated 0.818 ± 0.056, 0.803 ± 0.063, and 0.792 ± 0.076, respectively [54]. Furthermore, two of the MIE targets, peroxisome proliferator activated receptor γ (PPARγ) agonist (PPARg_ago, AID:743140) and aromatase antagonist (Arom_ant, AID:743139), improved the prediction-performance by optimizing of parameters in the modified DeepSnap-DL system, such as angle in the depiction of the image from 3D-chemicals, data-split ratio with training (train), validation (valid), and test datasets, background color in an image, and learning rate (LR) and batch size (BS) in hyperparameters in DL [54].…”

“…In the modified DeepSnap-DL system, the mean values of receiver operating characteristic area under the curve (ROC_AUC) of the prediction models for 59 MIE targets in validation, test, and foldout datasets indicated 0.818 ± 0.056, 0.803 ± 0.063, and 0.792 ± 0.076, respectively [54]. Furthermore, two of the MIE targets, peroxisome proliferator activated receptor γ (PPARγ) agonist (PPARg_ago, AID:743140) and aromatase antagonist (Arom_ant, AID:743139), improved the prediction-performance by optimizing of parameters in the modified DeepSnap-DL system, such as angle in the depiction of the image from 3D-chemicals, data-split ratio with training (train), validation (valid), and test datasets, background color in an image, and learning rate (LR) and batch size (BS) in hyperparameters in DL [54].…”