Automated High-Throuput Estimation and Modeling of Protein Network Distributions

描述（由申请人提供）：本研究旨在开发和测试方法，这些方法将使形成网络（如微管）的蛋白质的亚细胞分布的蛋白质组范围内的高通量研究成为可能，并用于识别与这些分布相关的蛋白质。更具体地说，我们将测试和进一步完善我们已经开发的方法，以估计大量细胞的细丝分布的平均特性（表征统计变化）。我们的目标是利用来自许多（可能是数千个）细胞的数据的可用性所提供的力量来估计这类蛋白质的高通量研究的核心问题：一些独立变量（药物或其他实验条件）对感兴趣的网络细丝分布的总体平均影响是什么。初步工作已经建立了这种方法的可行性，我们建议广泛测试它与真实的数据微管在各种细胞类型的各种条件下，扰动分布在已知的方式。结果将确定单个微管生长模型（具有参数变化）是否对许多细胞类型有效（即，以消除仅仅由于单元尺寸和形状引起的变化）。我们还将扩展这种方法，以确定许多未知蛋白质与细丝网络的相关性（亲和力），用于目前产生此类数据的几项蛋白质组研究。这些方法将用于分析来自现有和正在进行的蛋白质组规模研究的数千种蛋白质的图像。将通过与现有蛋白质数据库和文献中的信息进行比较以及通过额外的实验，对选定的示例测试仅基于可能与微管相关的特定蛋白质的基于图像的建模的识别。这项研究的成功完成不仅将提供有关许多蛋白质位置的重要新信息，而且将填补目前蛋白质组研究建模方法的空白，并促进高通量筛选实验中不同药物，siRNA或突变的作用的机制量化。 公共卫生相关性：本研究旨在通过生成建模方法对亚细胞毒性型蛋白质结构和分布进行基本了解。这项研究的成功完成将使类似的蛋白质和亚细胞结构的建模方法的发展，并促进不同的药物，siRNA或突变的高通量筛选实验中的作用的机制量化。

DESCRIPTION (provided by applicant): This research is aimed at developing and testing methods that will enable proteome-wide high-throughput studies of the subcellular distributions of proteins that form networks (such as microtubules), and for identifying proteins whose distributions are related to these. More specifically we will test and further refine methods that we have developed to estimate the average properties (characterize the statistical variation) of the filament distributions for a large number of cells. Our aim is to use the power afforded by the availability of data arising from many (possibly thousands) of cells to estimate what is really at the heart of the question for high-throughput studies of proteins of this type: what is the overall average effect of some independent variable (drug or other experimental condition) on the network filament distribution of interest. Preliminary work has established the feasibility of this approach, and we propose to test it extensively with real data for microtubules in a variety of cell types under a variety of conditions that perturb distributions in known ways. The results will establish whether a single microtubule growth model (with changes in parameters) is valid for many cell types (i.e., to remove variation due solely to cell size and shape). We will also extend this method to determine the correlation (affinity) of many unknown proteins to filament networks for several proteome wide studies currently generating such data. The methods will be used to analyze images for thousands of proteins from existing and ongoing proteome-scale studies. The identification solely on the basis of image-based modeling of specific proteins as likely to be microtubule-associated will be tested for selected examples by comparison with information in existing protein databases and literature and by additional experimentation. The successful completion of this study would not only provide important new information about the location of many proteins, but will fill a current void in modeling approaches for proteome-wide studies and facilitate the mechanistic quantification of effects of different drugs, siRNAs or mutations in high throughput screening experiments. PUBLIC HEALTH RELEVANCE: This research is aimed at enabling fundamental understanding of subcellular filament-type protein structure and distribution through generative modeling approaches. The successful completion of this study would enable the development of modeling approaches for similar proteins and subcellular structures and facilitate the mechanistic quantification of effects of different drugs, siRNAs or mutations in high throughput screening experiments.