调查是否深度学习模型的共同折叠学习蛋白质-连接体相互作用的物理
Matthew R Masters1,2, Amr H Mahmoud1,2, Markus A Lill3,4
1Department of Pharmaceutical Sciences, University of Basel, Basel, Switzerland.
Nature communications
|October 6, 2025
概括
对于蛋白质连接体结构预测的深度学习模型显示出局限性. 对抗性测试揭示了与物理原理的差异,暗示了潜在的过拟合和泛化问题.
科学领域:
- 计算生物学 计算生物学
- 结构生物学 结构生物学
- 人工智能的人工智能
背景情况:
- 像AlphaFold3和RoseTTAFold All-Atom这样的深度学习模型在蛋白质-连接体结构预测方面表现出色.
- 这些模型在药物发现和蛋白质工程方面具有广泛的潜力.
研究的目的:
- 批判性地评估先进的共同折叠模型对基本物理原理的坚持.
- 确定这些深度学习模型的概括能力的局限性.
主要方法:
- 利用基于既定的物理,化学和生物原理的对抗性示例.
- 在生物和化学上对蛋白质-连接体结构进行了可信的干扰.
主要成果:
- 在扰动下预测的蛋白质-连接体结构中显示了显著的差异.
- 观察到与预期的身体行为有显著差异,表明潜在的过拟合.
- 突出了对各种蛋白质-连接体相互作用的概括的局限性.
结论:
- 同折叠模型可能过度适应训练数据,损害了对物理定律的遵守.
- 强大的物理和化学先验对于提高这些预测工具的可靠性至关重要.
- 对于药物发现和蛋白质工程等关键应用,建议采取谨慎的方法.
更多相关视频
10:21Author Spotlight: Streamlining Protein Target Prediction and Validation via Molecular Docking and CETSA
Published on: February 23, 2024
3.6K
09:17Structure-Based Simulation and Sampling of Transcription Factor Protein Movements along DNA from Atomic-Scale Stepping to Coarse-Grained Diffusion
Published on: March 1, 2022
3.5K
相关概念视频
Protein Folding
126.3K
Overview
126.3K
Protein Folding
11.1K
Proteins are chains of amino acids linked together by peptide bonds. Upon synthesis, a protein folds into a three-dimensional conformation, critical to its biological function. Interactions between its constituent amino acids guide protein folding, and hence the protein structure is primarily dependent on its amino acid sequence.
Protein Structure Is Critical to Its Biological Function
Proteins perform a wide range of biological functions such as catalyzing chemical reactions, providing...
Protein Structure Is Critical to Its Biological Function
Proteins perform a wide range of biological functions such as catalyzing chemical reactions, providing...
11.1K
Ligand Binding Sites
14.9K
Proteins are dynamic macromolecules that carry out a wide variety of essential processes; however, the activities of most proteins depend on their interactions with other molecules or ions, known as ligands.
Protein-ligand interactions are quite specific; even though numerous potential ligands surround a cellular protein at any given time, only a particular ligand can bind to that protein. Moreover, a ligand binds only to a dedicated area on the surface of the protein, known as the...
Protein-ligand interactions are quite specific; even though numerous potential ligands surround a cellular protein at any given time, only a particular ligand can bind to that protein. Moreover, a ligand binds only to a dedicated area on the surface of the protein, known as the...
14.9K
Ligand Binding Sites
8.6K
8.6K
Noncovalent Attractions in Biomolecules
63.2K
Noncovalent attractions are associations within and between molecules that influence the shape and structural stability of complexes. These interactions differ from covalent bonding in that they do not involve sharing of electrons.
Four types of noncovalent interactions are hydrogen bonds, van der Waals forces, ionic bonds, and hydrophobic interactions.
Hydrogen bonding results from the electrostatic attraction of a hydrogen atom covalently bonded to a strong-electronegative atom like oxygen,...
Four types of noncovalent interactions are hydrogen bonds, van der Waals forces, ionic bonds, and hydrophobic interactions.
Hydrogen bonding results from the electrostatic attraction of a hydrogen atom covalently bonded to a strong-electronegative atom like oxygen,...
63.2K
Cooperative Allosteric Transitions
2.6K
2.6K
