ROBUST PORTABLE SOFTWARE FOR LOCATION PROTEOMICS

描述（由申请人提供）： 系统了解蛋白质的亚细胞位置以及这些位置在细胞周期、发育和疾病期间对药物的反应如何变化，对于当前全面的蛋白质组学工作至关重要。处理这一主题的新领域，定位蛋白质组学，需要荧光标记大量蛋白质的方法，快速收集高分辨率荧光显微镜图像，并自动分析结果分布。Murphy小组先前描述了自动化系统，该系统可以识别2D和3D图像中的所有主要亚细胞模式，并且可以比视觉检查更好地区分蛋白质模式。这些系统的一个关键组成部分是描述每个亚细胞模式的数字特征集，而不会对细胞大小和形状的变化过于敏感。本提案的第一个组成部分是研究开发和测试新分析问题的解决方案。在这一目标下，第一个目标是建立一个最佳的蛋白质分层分组，使其模式在所有条件下统计学上无法区分的蛋白质被放置在同一组中。第二种方法是不仅根据图像的静态模式，而且根据这些模式随时间的变化来描述和分类图像。第三是实现基于蛋白质位置模式相似性的分布式数据库图像快速检索方法。第二个组成部分是将目前的“研究级”软件转换为“分发级”软件，以便研究人员既可以将其作为独立的应用软件，也可以将其作为图像数据库系统的一部分使用。这些包括图像集的模式分类器和比较器。项目组包括多媒体数据库检索和数据挖掘以及软件工程原理方面的领先专家。开发的软件不仅对大规模蛋白质组学研究有用，而且对传统细胞生物学实验的自动化解释也有用。蛋白质按位置的分层分组，结合它们的序列，将首次允许预测位置的程序以发现新靶向基序所需的高分辨率运行。该项目实现的定位蛋白质组学的综合方法有望识别与疾病相关的蛋白质差异，这些差异可用作治疗靶点或作为早期检测和异常分类的标志物。

DESCRIPTION (provided by applicant): Systematic knowledge of the subcellular locations of proteins and how these locations change in response to drugs, during the cell cycle, during development and during disease will be essential to current comprehensive proteomics efforts. The new field dealing with this subject, Location Proteomics, requires methods for fluorescently-tagging large numbers of proteins, rapidly collecting high-resolution fluorescence microscope images, and, critically, automatically analyzing the resulting distributions. The Murphy group has previously described automated systems that can recognize all major subcellular patterns in 2D and 3D images and can distinguish protein patterns better than visual examination. A critical component of these systems is sets of numerical features that describe each subcellular pattern without being overly sensitive to variations in cell size and shape. The first component of this proposal is research to develop and test solutions to new analysis problems. Within this goal, the first aim is the building of an optimal hierarchical grouping of proteins such that proteins whose patterns are statistically indistinguishable under all conditions are placed in the same group. The second is describing and classifying images based not only on their static patterns but on how those patterns change over time. The third is implementing fast methods for retrieving images from distributed databases based on similarity of protein location patterns. The second component is conversion of current "research-grade" software to "distribution-grade" so that it can be made available to researchers both as stand-alone applications and as part of an image database system. These include pattern classifiers and comparators for image sets. The project team includes leading experts in multimedia database retrieval and data mining and software engineering principles. The software to be developed will be useful not only for large-scale proteomics efforts but also for automated interpretation of traditional cell biology experiments. The hierarchical grouping of proteins by location, in combination with their sequences, will for the first time allow programs for predicting location to operate at the high resolution necessary for discovery of new targeting motifs. The comprehensive approach to location proteomics enabled by this project promises to identify protein differences associated with disease, which can be used as targets for therapies or as markers for early detection and classification of abnormalities.