Scholarlink

<p>The drug target affinity (DTA) identification plays a major role in analyzing drug repositioning and drug mechanisms. The conventional machine learning (ML) approaches perform DTA identification as the binary classification model. This work presents an optimization based deep learning (DL) approach for automated DTA prediction. The process is carried out in five stages: dataset acquisition, pre-processing, data balancing, feature extraction and optimal classification. Initially, the datasets are acquired and the feature scaling process is used to normalize the data features. Then, for data balancing, the algorithm SMOTE with clustering is presented. Finally, the 1D convolutional self attention based long short time memory (1DCNN-SA-LSTM) is used for DTA prediction. Further, to optimize the hyperparameters of the model, Honey badger optimization (HBO) is introduced. Experimental results have shown that optimized 1DCNN-SA-LSTM technique outperforms the conventional models.</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

An optimized automated drug-target interaction identification model based on a dual approach of 1D convolutional neural networks and self-attention-enhanced LSTMs

  • 期刊论文

  • 发表日期：

    2025年1月23日

查看全文
  • Tintu Vijayan,
  • Pamela Vinitha Eric

摘要

The drug target affinity (DTA) identification plays a major role in analyzing drug repositioning and drug mechanisms. The conventional machine learning (ML) approaches perform DTA identification as the binary classification model. This work presents an optimization based deep learning (DL) approach for automated DTA prediction. The process is carried out in five stages: dataset acquisition, pre-processing, data balancing, feature extraction and optimal classification. Initially, the datasets are acquired and the feature scaling process is used to normalize the data features. Then, for data balancing, the algorithm SMOTE with clustering is presented. Finally, the 1D convolutional self attention based long short time memory (1DCNN-SA-LSTM) is used for DTA prediction. Further, to optimize the hyperparameters of the model, Honey badger optimization (HBO) is introduced. Experimental results have shown that optimized 1DCNN-SA-LSTM technique outperforms the conventional models.