南加州大学（University of Southern California）和华盛顿大学（University of Washington）的研究人员近日提出了一种新的深度学习模型——深度结合特异性预测器（DeepPBS），旨在通过几何深度学习技术预测蛋白质与DNA的结合特异性。这项研究为理解基因调控机制提供了新的工具，并在《Nature Methods》杂志上发表了相关成果。
蛋白质与DNA的相互作用对于基因的表达调控至关重要。然而，由于这些复合物的复杂性和多样性，预测蛋白质对特定DNA序列的结合特异性一直是一个挑战。DeepPBS模型利用几何深度学习，能够从蛋白质-DNA的结构中提取出可解释的结合特异性信息。
DeepPBS模型通过分析蛋白质-DNA的结构，能够提供每个蛋白质残基对DNA序列结合特异性的重要性评分。这些评分在蛋白质残基水平上聚合，并通过诱变实验得到验证。该模型不仅能够应用于已知的蛋白质-DNA结合结构，还能够预测没有实验结构数据的蛋白质的结合特异性。
研究显示，DeepPBS模型能够准确预测蛋白质-DNA复合物的结合特异性，并且不受蛋白质家族的限制，具有更广泛的适用性。此外，DeepPBS模型还可以与蛋白质结构预测方法结合使用，为蛋白质-DNA复合物的设计提供指导。
这项研究的成果对于生物信息学和分子生物学领域具有重要意义，它不仅能够促进对基因调控机制的理解，还可能为药物设计和基因治疗提供新的策略。随着生物技术的不断进步，DeepPBS模型有望在未来的生命科学研究中发挥更大的作用。
英语如下：
Title: USC Team Develops New Algorithm to Predict Specificity of Protein-DNA Binding
Keywords: Protein-DNA, Geometric Deep Learning, Specificity Prediction
News Content:
A team from the University of Southern California (USC) and the University of Washington has developed a new method to predict the specificity of protein-DNA interactions. The researchers have proposed a novel deep learning model, DeepPBS, which utilizes geometric deep learning techniques to predict the specificity of protein-DNA binding. This study provides a new tool for understanding the mechanisms of gene regulation and has been published in the journal Nature Methods.
The interaction between proteins and DNA is crucial for the regulation of gene expression. However, due to the complexity and diversity of these complexes, predicting the specificity of protein binding to specific DNA sequences has been a challenge. DeepPBS model utilizes geometric deep learning to extract interpretable binding specificity information from the structure of protein-DNA complexes.
The DeepPBS model can provide a scoring of the importance of each protein residue in the specificity of DNA sequence binding by analyzing the structure of protein-DNA complexes. These scores are aggregated at the protein residue level and validated through mutagenesis experiments. The model can be applied not only to known protein-DNA binding structures but also to predict the specificity of protein binding without experimental structural data.
The study demonstrates that the DeepPBS model can accurately predict the specificity of protein-DNA complex binding and is not limited by protein families, making it more widely applicable. Additionally, the DeepPBS model can be used in conjunction with protein structure prediction methods to guide the design of protein-DNA complexes.
The findings of this research are significant for the fields of bioinformatics and molecular biology, as they not only promote a better understanding of gene regulatory mechanisms but also potentially offer new strategies for drug design and gene therapy. As biotechnology continues to advance, the DeepPBS model is expected to play a greater role in future life science research.
【来源】https://www.jiqizhixin.com/articles/2024-08-19
Views: 1