结合蛋白水平的多组学数据整合识别基因功能及致病基因

发布时间：2018-03-22 11:04

本文选题：遗传疾病　切入点：功能相似性　出处：《南京航空航天大学》2017年硕士论文　论文类型：学位论文

【摘要】：致病基因预测一直是生物信息学领域的一个核心问题,发现致病基因是理解疾病发生机制、协助临床判断和预防治疗的基础。近年来产生了越来越多的基于功能相似性的致病基因预测方法,然而在目前定位的26000多个基因中,约有42%的基因的功能是未知的,功能注释的缺失会导致一些疾病候选基因在预测中被忽视掉。而如果仅仅在基因水平对基因功能进行预测是片面的,作为生物体内各种功能的主要执行者,蛋白质表达与基因功能息息相关,这就需要结合蛋白水平数据研究基因的功能。基于此,本文将结合人类蛋白质组表达数据并整合其他多组学数据预测人类未知基因的功能,预测致病基因。本文主要工作内容如下:本文提出了新的基因功能预测方法Pemo,其基于结合了蛋白质表达量的多组学数据预测基因功能。首先分析通过质谱实验得到的人类蛋白质表达量数据,构建未知功能基因的相关系数矩阵,利用相互作用网络筛选假阳性结果,根据组织特异性和条件概率构建先验概率功能矩阵,通过两个矩阵的乘积对功能注释进行打分排序。接着利用序列信息,RNA-Seq数据和互作网络数据分别对未知基因的功能进行预测,最后Pemo整合这四种组学数据对基因注释。在此基础上,本文结合蛋白水平信息和功能注释计算候选基因与遗传疾病之间的相关性。首先计算遗传疾病与Gene Ontology(GO)注释之间的关系,然后充分挖掘GO数据库功能注释之间的关系,综合考虑注释在GO结构中的距离和其父节点的交集,同时结合蛋白质表达量建立合理的评价机制,评估遗传疾病与基因的关系。在基因功能预测方面,本文首先比较了多种组学数据的预测结果,其中基于蛋白质表达数据集的方法在预测准确率上明显高于其他组学数据。然后分析了基于单一组学数据的预测和整合多组学数据的预测结果,证明Pemo整合多组学数据的策略是有效的,在结合了蛋白水平的信息后,预测准确率有了很大提高。比较Pemo方法和其他基因功能预测方法,在GO的三个方面Pemo的表现都是最好的。在致病基因预测方面,本文成功预测出胃癌、肺癌、乳腺癌和先天性心脏病等遗传疾病的致病基因,假阳性率也较低,在与其他致病基因预测方法进行比较时,本文方法的结果也是有优势的。在致病候选基因的相互作用网络图谱中,大多数致病基因参与共同的致病生物通路,其中一些参与程度较高的候选基因可能是潜在致病基因,有待今后实验进一步验证。
[Abstract]:The prediction of pathogenic genes has always been a core issue in the field of bioinformatics. In recent years, more and more genetic prediction methods based on functional similarity have been developed. However, about 42% of the more than 26000 genes currently located are unknown. The absence of functional annotation can cause some disease candidate genes to be ignored in the prediction. If it is one-sided to predict gene function only at the gene level, it is the main executor of various functions in the organism. Protein expression is closely related to gene function, which requires the study of gene function with protein level data. Based on this, this paper will combine human proteome expression data and integrate other sets of data to predict the function of unknown human genes. The main work of this paper is as follows: in this paper, a novel gene function prediction method, Pemom, is proposed, which is based on multigroup data combined with protein expression. Human protein expression data, The correlation coefficient matrix of unknown functional genes was constructed, the false positive results were screened by interaction network, and a priori probability function matrix was constructed according to tissue specificity and conditional probability. Then the function of unknown genes was predicted by sequence information RNA-Seq data and interaction network data. Finally, Pemo integrates these four genomes to annotate genes. On this basis, the relationship between candidate genes and genetic diseases is calculated with protein level information and functional annotation. The relationship between genetic diseases and Gene ontology o) annotations is first calculated. Then fully mining the relationship between the functional annotations of go database, considering the distance of annotation in go structure and the intersection of its parent nodes, and establishing a reasonable evaluation mechanism combined with protein expression. To evaluate the relationship between genetic diseases and genes. In the prediction of gene function, we first compared the predicted results of a variety of cluster data. The prediction accuracy of protein expression dataset is obviously higher than that of other sets of data. Then, the prediction results based on a single set of data and the integration of multiple sets of data are analyzed. It is proved that the strategy of integrating multiple sets of data into Pemo is effective, and the prediction accuracy has been greatly improved by combining the information of protein level. Comparing Pemo method with other gene function prediction methods, In the prediction of pathogenicity genes, we successfully predicted the pathogenetic genes of gastric cancer, lung cancer, breast cancer and congenital heart disease, and the false positive rate was also low. The results of this method are also superior when compared with other pathogenetic gene prediction methods. In the interaction network map of candidate genes, most of the pathogenic genes are involved in common pathogenicity pathways. Some of these candidate genes may be potential pathogenic genes, which need to be further verified by future experiments.
【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R440

【相似文献】