Scholarlink

<p>Advances in artificial intelligence, particularly in deep learning, are transforming the field of protein structure prediction. AlphaFold has emerged as a benchmark tool due to its remarkable ability to model protein folding based solely on amino acid sequences. Nevertheless, despite these impressive achievements, several prediction failures still occur. In this study, we focus on two multidomain proteins, LicT from <i>Bacillus subtilis</i> and P1 from the Rice Yellow Mottle Virus, that have long resisted structural characterization. Both proteins exhibit inter-domain flexibility, dimerization, and dynamic behavior in response to phosphorylation (for LicT) or zinc binding (for P1). Here, we screen the performances of AlphaFold versions against these two proteins endowed with a complex conformational landscape and show the present impossibility to correctly predict all domains simultaneously.</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

Challenging cases for AlphaFold: two multidomain proteins with zinc-binding-, phosphorylation- or dimerization-driven conformational changes

  • 期刊论文

  • 发表日期：

    2026年6月29日

查看全文
  • Yinshan Yang,
  • Hélène Déméné

摘要

Advances in artificial intelligence, particularly in deep learning, are transforming the field of protein structure prediction. AlphaFold has emerged as a benchmark tool due to its remarkable ability to model protein folding based solely on amino acid sequences. Nevertheless, despite these impressive achievements, several prediction failures still occur. In this study, we focus on two multidomain proteins, LicT from Bacillus subtilis and P1 from the Rice Yellow Mottle Virus, that have long resisted structural characterization. Both proteins exhibit inter-domain flexibility, dimerization, and dynamic behavior in response to phosphorylation (for LicT) or zinc binding (for P1). Here, we screen the performances of AlphaFold versions against these two proteins endowed with a complex conformational landscape and show the present impossibility to correctly predict all domains simultaneously.