A New Paradigm for Integrated Analysis of Multiscale Genomic Imaging Datasets

描述(由申请人提供)： 十年前，当微阵列第一次被发明时，它在自然遗传学中被誉为“希望的阵列”，并在生物医学界受到了相当大的关注。随后，它在《自然评论》杂志上被称为“一系列问题”。微阵列基因表达的一个固有问题是结构信息的缺失，这限制了它在生物发现中的能力。为了克服微阵列成像的低重复性和准确性，需要将基本范式转变为能够将互补和多尺度结构成像信息纳入微阵列成像的范式。幸运的是，高分辨率生物分子成像探针的最新进展加上先进的图像分析使细胞系统的综合和系统研究成为可能。可以使用多尺度和多模式成像来标记细胞，提供结构和功能信息。随着多尺度成像电子表格的出现，生命科学界迫切需要有效地管理这些信息，对其进行全面分析，并将由此产生的知识应用于理解细胞的遗传系统。然而，对这种大规模成像信息的管理和挖掘受到当今计算方法和知识共享基础设施的限制。这些问题是生物分子图像信息学新兴领域取得进展的主要障碍。因此，该项目的目标是开发一种独特的基因组图像管理和挖掘系统，使遗传学家能够以易于操作的方式搜索、关联和整合这些多尺度和多模式的成像信息，并进一步实现新的生物学发现。特别是，该系统将填补目前诸如开放显微镜环境(OME)等图像数据库系统中留下的空白，例如缺乏用于综合数据分析的分析工具。为了实现这一目标，我们正在组建一支由成像工程师、遗传学家和工业成像科学家组成的强大的跨学科团队。在我们多样化和互补性专业知识的基础上，我们能够提供创新和跨学科的方法，将图像处理、图像数据库设计和机器学习的最新进展与基因组学中高分辨率和高通量的分子成像探针的开发结合在一起。具体地说，我们将实现以下具体目标。首先，我们将开发一套从高分辨率荧光原位杂交(FISH)图像中提取内容和信息检索的算法。这个视觉系统将有效地管理成像表型信息，促进知识发现，如识别视觉上相似的亚型。其次，我们将把从FISH成像中提取的数量性状与基因组结构重排和基因表达模式相关联。最后，我们将开发一种数据集成方法来融合来自多模式成像数据库的不同信息，以改进生物系统的表征。

DESCRIPTION (provided by applicant): A decade ago when microarray was first invented, it was hailed as "an array of hope" in Nature Genetics and has received a considerable amount of attention in biomedicine. Subsequently it has been called "an array of problems" in Nature Review. An inherent problem with microarray gene expression is that structural information is missing, which limits its ability in biological discovery. To overcome the poor reproducibility and accuracy of microarray imaging, there needs to be a shift in fundamental paradigms to those able to incorporate complementary and multiscale structural imaging information into microarray imaging. Fortunately, the latest progress in high resolution biomolecular imaging probe development coupled with advanced image analysis makes integrative and systematic studies of cellular systems possible. A cell can be labeled using multiscale and multimodality imaging, providing both structural and functional information. With multiscale imaging spreadsheets now available, there is an overwhelming need within the life sciences community to manage this information effectively, to analyze it comprehensively, and to apply the resulting knowledge in the understanding of the genetic system of a cell. However, the management and mining of this large-scale imaging information is limited by today's computational approaches and knowledge-sharing infrastructure. These problems represent a major impediment to progress in the emerging area of bio-molecular image informatics. Therefore, the goal of this project is to develop a unique genomic image management and mining system that can allow geneticists to search, correlate and integrate this multiscale and multi-modality imaging information in an easily operable fashion and further enable new biological discovery. In particular, this system will fill a void left in the current image database systems such as Open Microscope Environment (OME), e.g., the lack of analytic tools for integrative data analysis. To realize this goal, we are bringing together a strong interdisciplinary team consisting of imaging engineers, geneticists and industrial imaging scientists. Building on our diverse and complementary expertise, we are able to provide innovative and interdisciplinary approaches that combine the latest progress in image processing, imaging database design and machine learning with the development of high resolution and high throughput molecular imaging probes in genomics. More specifically, we will accomplish the following specific aims. First, we will develop a suite of algorithms for content extraction and information retrieval from high resolution fluorescence in situ hybridization (FISH) images. This visual system will effectively manage imaging phenotype information, facilitating knowledge discovery such as identifying visually similar subtypes. Second, we will correlate quantitative traits extracted from FISH imaging with genomic structural rearrangements and gene expression patterns. Finally, we will develop a data integration approach to fuse disparate information from multi-modality imaging databases for improved characterization of biological systems.