使用分子模拟和基于物理的多重机器学习策略预测混乱蛋白质的物理特征
Diego Linares Gonzalez1, Shahana Ibrahim1, Swarnadeep Seth2
1Department of Electrical and Computer Engineering, University of Central Florida, Orlando, Florida 32816-2385, United States.
Biomacromolecules
|October 14, 2025
概括
我们开发了一种混合机器学习 (ML) 模型,以预测内在无序蛋白质 (IDP) 的结构性质. 我们的注意力引导框架整合了序列和物理特征,提高了预测准确性和可解释性,以进行高效的国内流离失所者查.
科学领域:
- 计算生物学 计算生物学
- 生物物理学的生物物理.
- 机器学习 机器学习
背景情况:
- 内在无序的蛋白质 (IDP) 缺乏稳定的3D结构,这对传统的蛋白质建模构成了挑战.
- 预测IDP的结构性质对于理解它们的生物功能至关重要.
研究的目的:
- 开发一种新的混合机器学习 (ML) 框架,用于准确预测 IDP 构造性质,例如旋转半径.
- 用注意力机制将序列信息与物理特征集成,以提高预测能力.
主要方法:
- 一个混合ML框架,将基于序列的模型 (例如GRU,biGRU) 与23个物理特征相结合.
- 一个注意力机制来衡量残留物的重要性,以及一个共同的潜在空间来进行特征融合.
- 在布朗动力学 (BD) 模拟数据上的培训和评估,用于来自MobiDB数据库的约7000名IDP.
主要成果:
- 混合型biGRU模型实现了最好的预测性能,超过了仅序列和仅特征模型.
- 注意引导的融合显著改善了准确度指标 (例如,平均绝对百分比误差,平均平方误差).
- SHAP和综合梯度分析确定了影响预测的关键特征 (例如,序列电荷,水分,不对称性) 和蛋白质长度.
结论:
- 开发的ML框架为预测IDP行为提供了一个快速,可解释和可扩展的工具.
- 该模型能够有效地对IDP进行初始查,减少了对广泛分子模拟的需求.
- 功能重要性分析有助于理解IDP的形状决定因素,并指导功能选择以改进概括.
相关概念视频
02:18Intrinsically Disordered Proteins
19.2K
Intrinsically disordered proteins are a group of proteins that do not fold into specific three-dimensional structures. Their structural flexibility allows them to complement ordered proteins to perform functions that are inaccessible to rigid structures. They are more common in eukaryotes than prokaryotes and may either be exclusively intrinsically disordered or hybrid proteins, consisting of a mix of ordered and disordered regions. The absence of a rigid structure in these proteins can be...
19.2K
02:18Intrinsically Disordered Proteins
2.8K
2.8K
01:49Conserved Binding Sites
5.0K
Many proteins’ biological role depends on their interactions with their ligands, small molecules that bind to specific locations on the protein known as ligand-binding sites. Ligand-binding sites are often conserved among homologous proteins as these sites are critical for protein function.
Binding sites are often located in large pockets, and if their location on a protein’s surface is unknown, it can be predicted using various approaches. The energetic method computationally...
Binding sites are often located in large pockets, and if their location on a protein’s surface is unknown, it can be predicted using various approaches. The energetic method computationally...
5.0K
01:24Protein Organization
9.1K
Proteins are polymers of amino acid residues. They are versatile and responsible for different cellular functions, including DNA replication, molecular transport, catalysis, and structural support. Proteins have a hierarchical structure comprising at least three levels of organization: primary, secondary, and tertiary structure. Some large proteins have a quaternary structure where individual protein subunits are linked together.
The primary structure of a protein is its amino acid sequence....
The primary structure of a protein is its amino acid sequence....
9.1K
01:33Proteomics
9.3K
A proteome is the entire set of proteins that a cell type produces. We can study proteomes using the knowledge of genomes because genes code for mRNAs, and the mRNAs encode proteins. Although mRNA analysis is a step in the right direction, not all mRNAs are translated into proteins.
Proteomics is the study of proteomes' function. It involves the large-scale systematic study of the proteome to denote the protein complement expressed by a genome. Scientist Mark Wilkins coined the term...
Proteomics is the study of proteomes' function. It involves the large-scale systematic study of the proteome to denote the protein complement expressed by a genome. Scientist Mark Wilkins coined the term...
9.3K
02:27Predicting Molecular Geometry
45.0K
VSEPR Theory for Determination of Electron Pair Geometries
45.0K