CAREER: Scalable Computational Models for Multicellular Systems Biology

我们以大约200种细胞类型的复合体进入生命，这些细胞类型被编排成一个后生动物有机体；几天内，数以万亿计的细菌、古生物和真核微生物加入到这些细胞中，这些微生物居住在我们身体的每一个表面。虽然几十年的微生物生态学和后生动物细胞生物学已经详细描述了这些实体的许多方面，但我们最近才开始在分子水平上将单细胞培养中的单细胞生物体模型与复杂的多细胞系统联系起来。因此，这项研究的目标是开发计算方法，通过利用大型实验数据库来模拟多细胞系统的分子行为，特别是微生物群落及其与后生动物组织的相互作用。该项目的重点是确定基因产品在这类系统中的生物学作用，并转换来自受控实验环境的数据，以便应用于多细胞类型和多物种的部分。这将需要开发数据挖掘算法，能够有效地利用来自不同生物体的数千个实验数据集来模拟多细胞生物学的关键方面：多物种群落、细胞类型和谱系，以及它们在群落或组织中的结构和分布。为此目的，该项目将开发用于机器学习的卫星模型，其中核心属性和参数可根据需要进行修改。这些模型的预测将通过对口腔和肠道微生物区系中的微生物及其功能活动的表征，以及对个别特征不足的微生物和微生物相互作用的表征，在实验上得到验证。已开发工具的开源和在线实施将通过http://huttenhower.sph.harvard.edu.This项目的实验室网站提供，该项目将为由多种物种或细胞类型组成的多细胞系统中的基因组数据挖掘提供一个通用框架，并具有一个简单的界面，用于汇总数以千计的基因组规模的数据集。教育部分将包括扩展定量基因组学计划，其中包括新开发的计算生物学课程，通过斯坦福大学南非生物医学信息学计划进行推广，以及与哈佛大学LS/HHMI和国际计算生物学高中推广计划的持续合作。这将为通过挖掘大型生物数据集来理解多细胞系统和相互作用的培训和计算方法奠定坚实的基础。

We enter life as a composite of some 200 cell types orchestrated into a single metazoan organism; within days, these are joined by trillions of bacterial, archaeal, and eukaryotic microbes resident on every surface of our bodies. While decades of microbial ecology and metazoan cellular biology have detailed many aspects of these entities, we have only recently begun to bridge models of unicellular organisms in monoculture with complex multicellular systems at the molecular level. The goal of this research is thus to develop computational methodology to model the molecular behavior of multicellular systems, particularly microbial communities and their interactions with metazoan tissues, by taking advantage of large experimental data repositories. The project focuses on characterizing the biological roles of gene products in such systems and on translating data from controlled experimental contexts so as to apply in multi-cell-type and multi-species moieties. This will require the development of data mining algorithms capable of efficiently leveraging thousands of experimental datasets from diverse organisms to model key aspects of multicellular biology: multi-species communities, cell types and lineages, and their structure and distribution within a community or tissue. For this purpose, the project will develop satellite models for machine learning in which core properties and parameters are modified on an as-needed basis. Predictions from these models will be experimentally validated by characterization of the organisms in and functional activity of the oral and gut microbiota and of individual under-characterized microbes and microbial interactions. Open-source and online implementations of developed tools will be available through the laboratory web site at http://huttenhower.sph.harvard.edu.This project will provide a general framework for genomic data mining in multicellular systems made up of multiple species or cell types, with a simple interface for summarizing thousands of genome-scale datasets. The educational component will include an expansion of the Program in Quantitative Genomics, which includes a newly-developed computational biology curriculum, outreach through the Stanford South Africa Biomedical Informatics program, and ongoing collaborations with the Harvard University LS/HHMI and International Society for Computational Biology high school outreach programs. This will establish solid foundations in training and in computational methodology for understanding multicellular systems and interactions by mining large biological data collections.