Scholarlink

<p>Molecular docking techniques utilizing computational methods play an important role nowadays in predicting drug-target interactions to minimize the time and expenses of the process. Quinoline carboxamide analogs has great interest due to their remarkable pharmacological and binding properties. Unfortunately, current molecular docking and machine learning approaches lack efficiency in modeling multi-scale molecular interactions and assigning the proper weight to crucial molecular features; therefore, making the predictions less accurate and reliable. This research aims at proposing a novel graph convolutional network model called the Attention-based Multi-Scale Fusion Graph Convolutional Network (MSFGCN) for predicting molecular docking of quinoline carboxamide analogs. This framework combines multi-scale graph convolution, node weighting using attention mechanism, adjacency reconstruction, scaffold-based split for data, and Principal Component Analysis (PCA) to improve molecular representation and model generalization. The results from experimentation indicate that the proposed MSFGCN performed better than Graph Attention Network (GAT), Random Forest, and Extreme Gradient Boosting (XGBoost), with AUC scores of 0.9920 on the training set and 0.9612 on the test set, respectively. Principal Component Analysis results show a similar chemical space distribution in the training set and test set. Docking study found that compound 2a showed highest binding affinity (-7.3&#xa0;kcal/mol). Furthermore, molecular dynamics analysis has shown improved structural stability due to decreased fluctuation, decreased solvent accessible surface area, and radius of gyration. The results show that the presented MSFGCN model is very accurate and can be used reliably for ligand docking prediction and proves useful in future for rapid drug discovery and computational drug design.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Attention based multi-scale fusion graphical framework for molecular docking of quinoline carboxamide analogues

  • 期刊论文

  • 发表日期：

    2026年7月25日

查看全文
  • P. Shunmugaraj,
  • T. F. Abbs Fen Reji

摘要

Molecular docking techniques utilizing computational methods play an important role nowadays in predicting drug-target interactions to minimize the time and expenses of the process. Quinoline carboxamide analogs has great interest due to their remarkable pharmacological and binding properties. Unfortunately, current molecular docking and machine learning approaches lack efficiency in modeling multi-scale molecular interactions and assigning the proper weight to crucial molecular features; therefore, making the predictions less accurate and reliable. This research aims at proposing a novel graph convolutional network model called the Attention-based Multi-Scale Fusion Graph Convolutional Network (MSFGCN) for predicting molecular docking of quinoline carboxamide analogs. This framework combines multi-scale graph convolution, node weighting using attention mechanism, adjacency reconstruction, scaffold-based split for data, and Principal Component Analysis (PCA) to improve molecular representation and model generalization. The results from experimentation indicate that the proposed MSFGCN performed better than Graph Attention Network (GAT), Random Forest, and Extreme Gradient Boosting (XGBoost), with AUC scores of 0.9920 on the training set and 0.9612 on the test set, respectively. Principal Component Analysis results show a similar chemical space distribution in the training set and test set. Docking study found that compound 2a showed highest binding affinity (-7.3 kcal/mol). Furthermore, molecular dynamics analysis has shown improved structural stability due to decreased fluctuation, decreased solvent accessible surface area, and radius of gyration. The results show that the presented MSFGCN model is very accurate and can be used reliably for ligand docking prediction and proves useful in future for rapid drug discovery and computational drug design.