Scholarlink

<p>Recent advances in deep learning have proven highly effective in medical applications, notably in drug discovery. Among various deep learning techniques, self-supervised learning (SSL) has shown considerable advantages over traditional supervised learning by utilizing vast amounts of unlabeled data for model training. This review discusses both classic and state-of-the-art SSL-based methods in the drug discovery field, detailing their applications from small molecule and peptide drug discovery to antibody design and vaccine development, which provides a current and accessible guide to drug discovery. Furthermore, this review suggests the challenges faced by SSL in drug discovery, such as data quality, model interpretability, and computational resource constraints, and outlines its potential future directions. As deep learning technology advances, we anticipate that SSL-based models will increasingly promote drug research and development, potentially revolutionizing the pharmaceutical industry.</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

Self-supervised learning in drug discovery

  • 期刊论文

  • 发表日期：

    2025年6月23日

查看全文
  • Yangyang Chen,
  • Zixu Wang,
  • Jianmin Wang,
  • Yanyi Chu,
  • Qingpeng Zhang,
  • Zhong Alan Li,
  • Xiangxiang Zeng

摘要

Recent advances in deep learning have proven highly effective in medical applications, notably in drug discovery. Among various deep learning techniques, self-supervised learning (SSL) has shown considerable advantages over traditional supervised learning by utilizing vast amounts of unlabeled data for model training. This review discusses both classic and state-of-the-art SSL-based methods in the drug discovery field, detailing their applications from small molecule and peptide drug discovery to antibody design and vaccine development, which provides a current and accessible guide to drug discovery. Furthermore, this review suggests the challenges faced by SSL in drug discovery, such as data quality, model interpretability, and computational resource constraints, and outlines its potential future directions. As deep learning technology advances, we anticipate that SSL-based models will increasingly promote drug research and development, potentially revolutionizing the pharmaceutical industry.