【作者】 钮俊清；

【导师】 王煦法；

【作者基本信息】 中国科学技术大学 ， 计算机应用技术， 2008， 博士

【摘要】 随着人类基因组计划的基本完成,生命科学研究进入了”后基因组”时代。在后基因组时代面临的一个重大挑战就是如何从整体层面上揭示生物系统中DNA、RNA、蛋白质和各种生物小分子通过相互作用而产生生命现象。在这一背景下产生了系统生物学,它是一门新兴交叉学科,其目的在于在系统层次上理解生物系统。由于生物系统的内在复杂性要成功地进行系统生物学研究,必须借助数学建模和计算机仿真的方法。建模和仿真生物体和细胞是非常困难的,其原因在于:首先,生物内在的复杂性和生物实验技术的限制,导致生物的知识和实验数据不足;其次,对于复杂的生物过程,发展新建模和仿真的方法来研究复杂的生物系统成为生物学面临的一个重要的挑战。针对当前的需求,本文提出多源数据整合与agent技术结合的方法来研究生化网络,主要研究内容包括:1.构建数据整合对于系统生物学研究是非常重要和有帮助的,本文提出一种以生化网络模型为中心的多源数据整合方法,基于此方法构建了一个面向具体生物问题的数据整合平台(BioDB)。它是面向生物具体的问题,围绕选定具体问题将相关的生物数据库进行整合。其次,它是以生物网络模型为中心来构建的数据整合系统,将相关的生物数据库、文献知识、专家知识、生物实验数据和仿真实验数据围绕生物模型来整合。实验显示BioDB为重构代谢网络提供一个有效的数据平台,使得重构不但拥有更好的结果,而且具有快速、高效的特点。2.针对生物数据标准无法共享应用的问题,本文提出一种将生物数据进行标准转换的方法(BioBridge),它为生物Pathway数据标准之间提供了一个稳定的桥梁,使得数据可以跨越标准进行共享和使用。数据联邦整合方法中数据访问的效率一直是研究的重要问题,我们提出一种基于有限记忆多LRU的web缓存替换算法来构建了基于web缓存的数据联邦系统(LinkDB),有效的提高了在线的获取数据的效率。3.现阶段对生物系统的建模和仿真技术和方法需要进一步的发展,大多数现有的方法致力于简单生物学过程的建模和仿真。本文提出了一种在分子尺度上基于agent的建模方法,基于此方法构建了一个基于agent技术的计算平台来分析生化网络。该方法通过研究agent行为自组织突生形成的复杂宏观现象,来揭示生物系统的内在机制和宏观复杂现象和微观分子行为之间的联系。4.通常的,实际的生物系统具有很高的复杂性,这给建模和仿真生化网络的方法提出了更高的计算要求。本文基于agent技术和并行化思想提出的一种分布式的随机仿真方法(DSSA),算法主要是通过将Gillespie的SSA算法有效的分解到基于多agent系统的分布式框架中,同时应用反应关系图来进一步的减少计算和通讯代价。实验显示DSSA算法在时间性能上带来很大的提升,特别是对于一些大型的生化网络系统。应用通过多源数据整合的基于agent建模的方法对于生化系统加以研究不仅具有巨大的理论价值,还具有广阔的应用前景,本文在通过多源数据整合来建模和仿真分析生化网络系统的体系下做了一些研究工作,如何进一步完善现有的方法和平台,研究生物系统内在的演进机制是我们未来的方向。更多还原

【Abstract】 With the human genome project’s basic completion, scientific research of life indicates that it has entered a "post-genomic" era. In the post-genome era facing a major challenge is how to reveal the phenomenon of life at the whole level, which arising from interactions among DNA, RNA, proteins and small molecules of various biological systems. Under this background systems biology are proposed, which is a new emerging interdisciplinary and its goal lies in understanding the biological system at system level. Because of inherent complexity of biological system to successfully carry out research of systems biology, we must use mathematical modeling and computer simulation methods for the inherent complexity of biological systems. Modeling and simulation of organisms and cells is very difficult. Several reasons are: Firstly, the inherent complexity of biological system and the constraints of the biological experiment technology, which causes the knowledge and the empirical datum is insufficient. Secondly, the study of molecular randomness of the biological systems and how the process from molecular behaviors in micro to the macro complex phenomenon become a huge challenge. This dissertation, the method of combining multi-source data integration and agent technology to study biochemical network, the main contents include:1. Data integration is very important and helps research on systems biology, this dissertation proposes a data integration method with biochemical network model as central and construct a data platform (BioDB), which faces to specific biological issues and integrates of biological databases related to the selected specific issues. We construct the data integration system with biochemical network model as central, and the others include the related biological databases, literature knowledge, expertise, experimental data and simulated data. Experiments showed that our BioDB provides an effective data platform for reconstruction of metabolic network, making reconstruction not only have better results, but also with rapid and efficient performances.2. For the problem of biological data standards cannot share their applications, this dissertation proposes a data conversion method among several biological stan-dards(BioBridge), which provide a bridge for several biological data standards and can share these standards and their applications. The efficiency of data access is an important issue of data federation. This dissertation proposes a limited history based multi-LRU web cache replacement algorithm and constructs a data federal system (LinkDB) with web cache, which effectively improving the efficiency of accessing data on-line.3. Mostly of the existing methods for modeling and simulation biological systems only fit to simple biological process, so modeling and simulation techniques and methods need further develop. In this dissertation, we propose an agent-based modeling method at molecular scale (ABMMS) and construct a computation platform based on agent technology to analyze biological networks. We can study their complex macro phenomenon emerge from behaviors of agents. It provides a new way to study and understand biological systems, which can reveal internal mechanisms of biological systems and the relation between complex macro phenomenon and molecular behaviors at micro.4. Usually, the actual biological systems with the very high complexity, we need the higher performance of computation for modeling and simulation biochemical network. Based on agent technology and parallel theory this dissertation proposes a new distributed-based stochastic simulation algorithms (DSSA) us(?) ing multi-agents system and distributed computing to improve computing performance SSA. DSSA mainly through decomposed SSA into the framework of based on distributed multi-agent system, and through reaction relationship to further reduce the cost of computing and communications. Experiments showed DSSA algorithm is able to improve time performance significantly, especially for some large-scale biochemical networks.The application of our method, studying the biochemical system through the multiple source data integration and agent-based modeling method, not only has great theoretical value, but also has broad application prospects. We has done some research work under the framework of study of biochemical network system by multi-source data integration and modeling and simulation. And our future directions are how to further improve the existing methods and platform system for the study of the evolution mechanism of biological systems.更多还原