Scholarlink

<p>Drug discovery is crucial for identifying therapeutic agents that effectively interact with disease-linked proteins. This interaction, quantified by drug-target binding affinity (DTBA), is a key indicator of a drug's potential efficacy. Traditional methods, like high-throughput screening, are resource-intensive and costly. While machine learning approaches have emerged for DTBA prediction, significant challenges remain. To address these challenges, we introduce TransMLP-DTBA, an innovative All-MLP Mixing based Transformer architecture for DTBA prediction. TransMLP-DTBA overcomes limitations of existing methods by avoiding computationally expensive operations like high-dimensional multiplications and Multihead-Attention. Instead, it employs MLP Mixing, reducing computational demands and streamlining the process, making it suitable for large-scale screenings and drug repurposing. TRansMLP-DTBA leverages 1D sequences of drugs and targets, represented by SMILES strings and protein sequences, respectively, bypassing the need for complex 3D structural data. Extensive testing on public datasets demonstrates TransMLP-DTBA superiority over state-of-the-art methods like DeepDTA and GraphDTA, achieving significantly lower mean square error, higher concordance index, and higher R-squared, highlighting its robustness and predictive accuracy. This research unlocks potential for further advancements in drug discovery, emphasizing the potential of MLP-Mixing based transformer models in biomedical applications.</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

Optimizing drug-target binding affinity prediction for kinase proteins: a novel all-MLP-transformer approach

  • 期刊论文

  • 发表日期：

    2025年4月16日

查看全文
  • Nassima Aleb

摘要

Drug discovery is crucial for identifying therapeutic agents that effectively interact with disease-linked proteins. This interaction, quantified by drug-target binding affinity (DTBA), is a key indicator of a drug's potential efficacy. Traditional methods, like high-throughput screening, are resource-intensive and costly. While machine learning approaches have emerged for DTBA prediction, significant challenges remain. To address these challenges, we introduce TransMLP-DTBA, an innovative All-MLP Mixing based Transformer architecture for DTBA prediction. TransMLP-DTBA overcomes limitations of existing methods by avoiding computationally expensive operations like high-dimensional multiplications and Multihead-Attention. Instead, it employs MLP Mixing, reducing computational demands and streamlining the process, making it suitable for large-scale screenings and drug repurposing. TRansMLP-DTBA leverages 1D sequences of drugs and targets, represented by SMILES strings and protein sequences, respectively, bypassing the need for complex 3D structural data. Extensive testing on public datasets demonstrates TransMLP-DTBA superiority over state-of-the-art methods like DeepDTA and GraphDTA, achieving significantly lower mean square error, higher concordance index, and higher R-squared, highlighting its robustness and predictive accuracy. This research unlocks potential for further advancements in drug discovery, emphasizing the potential of MLP-Mixing based transformer models in biomedical applications.