A UNIFIED MULTITASK ARCHITECTURE FOR PREDICTING LOCAL PROTEIN PROPERTIES

用于预测局部蛋白质特性的统一多任务架构

• 批准号: 8365897
• 负责人: William Noble
• 金额: $ 2.14万
• 依托单位: UNIVERSITY OF WASHINGTON
• 依托单位国家: 美国
• 财政年份: 2011
• 资助国家: 美国
• 起止时间: 2011-09-01 至 2012-06-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/8365897
• 关键词: Amino Acid Sequence; Amino Acids; Architecture; Biological Neural Networks; Biology; Computational Biology; Computer Architectures; DNA Binding; Dependency; Engineering; Funding; Fungal Genome; Grant; Joints; Label; Learning; Modeling; National Center for Research Resources; Natural Language Processing; Output; Pattern; Peptide Sequence Determination; Peptide Signal Sequences; Performance; Principal Investigator; Property; Proteins; Relative (related person); Research; Research Infrastructure; Resources; Solvents; Source; Structure; Time; Training; United States National Institutes of Health; Work; cost; multitask; novel; synthetic protein

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. Primary support for the subproject and the subproject's principal investigator may have been provided by other sources, including other NIH sources. The Total Cost listed for the subproject likely represents the estimated amount of Center infrastructure utilized by the subproject, not direct funding provided by the NCRR grant to the subproject or subproject staff. A variety of functionally important protein properties, such as secondary structure, transmembrane topology and solvent accessibility, can be encoded as a labeling of amino acids. Indeed, the prediction of such properties from the primary amino acid sequence is one of the core projects of computational biology. Accordingly, a panoply of approaches have been developed for predicting such properties; however, most such approaches focus on solving a single task at a time. Motivated by recent, successful work in natural language processing, we propose to use multitask learning to train a single, joint model that exploits the dependencies among these various labeling tasks. We describe a deep neural network architecture that, given a protein sequence, outputs a host of predicted local properties, including secondary structure, solvent accessibility, transmembrane topology, signal peptides and DNA-binding residues. The network is trained jointly on all these tasks in a supervised fashion, augmented with a novel form of semi-supervised learning in which the model is trained to distinguish between local patterns from natural and synthetic protein sequences. The task-independent architecture of the network obviates the need for task-specific feature engineering. We demonstrate that, for all of the tasks that we considered, our approach leads to statistically significant improvements in performance, relative to a single task neural network approach, and that the resulting model achieves state-of-the-art performance.

该子项目是利用资源的众多研究子项目之一 由 NIH/NCRR 资助的中心拨款提供。子项目的主要支持 并且子项目的主要研究者可能是由其他来源提供的， 包括其他 NIH 来源。 子项目可能列出的总成本 代表子项目使用的中心基础设施的估计数量， NCRR 赠款不直接向子项目或子项目工作人员提供资金。 多种具有重要功能的蛋白质特性，例如 二级结构、跨膜拓扑和溶剂可及性， 可以被编码为氨基酸的标记。 确实，预测 一级氨基酸序列的这些特性之一是 计算生物学的核心项目。 因此，一整套 已经开发出预测此类特性的方法； 然而，大多数此类方法侧重于一次性解决单个任务 时间。 受到最近自然语言领域成功工作的激励 处理，我们建议使用多任务学习来训练 单一的联合模型，利用这些不同的之间的依赖关系 标记任务。 我们描述了深度神经网络架构 给定一个蛋白质序列，输出一系列预测的局部 性质，包括二级结构、溶剂可及性、 跨膜拓扑、信号肽和 DNA 结合残基。 这 网络以监督方式针对所有这些任务进行联合训练， 增强了一种新颖的半监督学习形式，其中 模型被训练来区分局部模式和自然模式 和合成蛋白质序列。任务无关的架构 网络消除了对特定于任务的功能的需要 工程。我们证明，对于我们所执行的所有任务 考虑到，我们的方法导致统计显着 相对于单任务神经网络的性能改进 方法，并且所得模型达到了最先进的水平 表现。