基于复杂网络的疾病基因预测的研究
发布时间：2018-10-13 12:06
【摘要】：近年来,识别疾病的相关基因成为生命科学领域富有挑战性的工作之一。传统的预测疾病基因的方法有连锁分析(Linkage Analysis)和关联研究(Association Study)。但是连锁分析方法只能定位染色体上的一段区域,这段区域包含几十个到几百个基因。同时关联研究也需要明确候选基因。各国研究人员陆续的提出对这段区域的候选基因进行进一步筛选的方法。 人类基因组计划(Human Genome Project, HGP)的完成和高通量生物技术的产生,我们获取了大规模的人类蛋白质交互作用数据(Protein-Protein Interaction, PPI)。有研究表明,在PPI网络上,具有较高拓扑重叠的蛋白质共属于一个生物功能模块或生物通路的可能性就越大。基于此,本文中,我们提出了基于PPI网络的对疾病候选基因进行预测的方法MTOMATOM。此方法结合多点拓扑重叠法(Multi-node Topology Overlap Measure, MTOM)和两点拓扑重叠法(Averaged Topology Overlap Measure, ATOM)。 MTOMATOM方法是通过衡量网络节点间拓扑重叠性大小来反映网络节点间的相似性,是一个更能反映生物意义的网络距离度量法。我们把该方法在包含783个基因的110类疾病-基因家族中进行50-fold留一法交叉验证,发现enrichment达到27-fold, roc曲线下面积为92.3%,取得了与同类方法相比较好的效果。 我们把MTOMATOM方法应用于阿尔茨海默氏病(Alzheimer's Disease,AD)相关基因的发现研究。首先,在kohler等人构建的PPI网络上进行预测分析,取得了跟kohler等人提出来的全局度量法随机游走相同的效果。其次,基于刘等人提出来的脑特异网络进行AD基因的预测,前46个分值最高的基因中,有40个与AD相关联的基因,比刘等人的预测结果稍好。MTOMATOM方法复杂度低,运算速度快,并且对网络的不完整性和连接的假阳性有较强的鲁棒性。
[Abstract]:In recent years, the identification of disease-related genes has become one of the challenging tasks in life sciences. The traditional methods for predicting disease genes are linkage analysis (Linkage Analysis) and association study (Association Study). But linkage analysis can only locate a region of a chromosome that contains dozens to hundreds of genes. At the same time, association studies also need to identify candidate genes. Researchers from all over the world have proposed methods for further screening candidate genes in this region. With the completion of the Human Genome Project (Human Genome Project, HGP) and the production of high-throughput biotechnology, we have obtained large-scale human protein interaction data (Protein-Protein Interaction, PPI). Studies have shown that on PPI networks, proteins with high topological overlaps are more likely to belong to one biological functional module or biological pathway. Therefore, in this paper, we propose a method of disease candidate gene prediction based on PPI network, MTOMATOM. This method combines multi-point topology overlap method (Multi-node Topology Overlap Measure, MTOM) and two-point topological overlap method (Averaged Topology Overlap Measure, ATOM). MTOMATOM method) to reflect the similarity of network nodes by measuring the degree of topological overlap between nodes. It is a network distance measure that can reflect biological meaning more. The method was cross-validated by 50-fold method in 110 disease-gene families containing 783 genes. It was found that the enrichment reached 27-fold and the area under the roc curve was 92.3.The results were better than that of the similar methods. We applied the MTOMATOM method to the discovery of genes associated with Alzheimer's disease (Alzheimer's Disease,AD). First, the prediction analysis is carried out on the PPI network constructed by kohler et al., and the results are the same as the global metric proposed by kohler et al. Secondly, based on the brain-specific network proposed by Liu et al., 40 of the first 46 genes with the highest score are associated with AD, which is slightly better than that of Liu et al. the MTOMATOM method is less complex and faster. And it has strong robustness to the network imperfection and false positive connection.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2009
【分类号】：R346
本文编号：2268556
[Abstract]:In recent years, the identification of disease-related genes has become one of the challenging tasks in life sciences. The traditional methods for predicting disease genes are linkage analysis (Linkage Analysis) and association study (Association Study). But linkage analysis can only locate a region of a chromosome that contains dozens to hundreds of genes. At the same time, association studies also need to identify candidate genes. Researchers from all over the world have proposed methods for further screening candidate genes in this region. With the completion of the Human Genome Project (Human Genome Project, HGP) and the production of high-throughput biotechnology, we have obtained large-scale human protein interaction data (Protein-Protein Interaction, PPI). Studies have shown that on PPI networks, proteins with high topological overlaps are more likely to belong to one biological functional module or biological pathway. Therefore, in this paper, we propose a method of disease candidate gene prediction based on PPI network, MTOMATOM. This method combines multi-point topology overlap method (Multi-node Topology Overlap Measure, MTOM) and two-point topological overlap method (Averaged Topology Overlap Measure, ATOM). MTOMATOM method) to reflect the similarity of network nodes by measuring the degree of topological overlap between nodes. It is a network distance measure that can reflect biological meaning more. The method was cross-validated by 50-fold method in 110 disease-gene families containing 783 genes. It was found that the enrichment reached 27-fold and the area under the roc curve was 92.3.The results were better than that of the similar methods. We applied the MTOMATOM method to the discovery of genes associated with Alzheimer's disease (Alzheimer's Disease,AD). First, the prediction analysis is carried out on the PPI network constructed by kohler et al., and the results are the same as the global metric proposed by kohler et al. Secondly, based on the brain-specific network proposed by Liu et al., 40 of the first 46 genes with the highest score are associated with AD, which is slightly better than that of Liu et al. the MTOMATOM method is less complex and faster. And it has strong robustness to the network imperfection and false positive connection.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2009
【分类号】：R346
【共引文献】
相关期刊论文 前2条
1 ;Biomarkers of Alzheimer’s disease in body fluids[J];Science China(Life Sciences);2010年04期
2 郑妍鹏;何金生;洪涛;;阿尔茨海默病体液生物学标记物研究进展[J];中国科学(C辑:生命科学);2009年09期
，本文编号：2268556
本文链接：https://www.wllwen.com/yixuelunwen/shiyanyixue/2268556.html
最近更新
教材专著
