本文關(guān)鍵詞： 基因預(yù)測(cè) 表型 人類表型本體 表型網(wǎng)絡(luò)　出處：《哈爾濱工業(yè)大學(xué)》2015年碩士論文　論文類型：學(xué)位論文

【摘要】：下一代基因測(cè)序技術(shù)加快了基因數(shù)據(jù)產(chǎn)生的速度,如此大的數(shù)據(jù)量,使得如今的難題從測(cè)序轉(zhuǎn)向如何有效的利用已經(jīng)產(chǎn)生的基因數(shù)據(jù)。表型是指受基因、環(huán)境等影響而在生物體上表現(xiàn)出來(lái)的特征,因而表型和基因以及疾病等有著密切的聯(lián)系。現(xiàn)如今表型學(xué)已經(jīng)是基因和疾病之間的重要紐帶,而對(duì)表型的研究已是預(yù)測(cè)和發(fā)現(xiàn)致病基因的重要手段。目前在基因、表型和疾病之間的關(guān)聯(lián)方面的研究方法主要有三種。一個(gè)是利用現(xiàn)有的生物醫(yī)學(xué)文獻(xiàn),用文本發(fā)掘等相關(guān)技術(shù)發(fā)掘生物醫(yī)學(xué)實(shí)體之間的關(guān)系。另外一種是利用已有的關(guān)系建立起表型、疾病、蛋白質(zhì)、基因等實(shí)體間的網(wǎng)絡(luò),在網(wǎng)絡(luò)中發(fā)現(xiàn)新的關(guān)系。最后一種是利用本體,如基因本體、表型本體等結(jié)構(gòu)化的知識(shí)系統(tǒng)來(lái)計(jì)算實(shí)體間的相似關(guān)系。它們之間各有優(yōu)缺點(diǎn)。研究基于本體的表型相似度計(jì)算方法能夠幫助預(yù)測(cè)病人的致病基因和疾病,充分利用本體的價(jià)值。本文主要利用人類表型本體(Human Phenotype Ontology,HPO)作為工具來(lái)研究基因和表型之間以及疾病和表型之間的相似性關(guān)系,進(jìn)而預(yù)測(cè)病人的致病基因和疾病。本文在基于人類表型本體中表型信息量的基礎(chǔ)上結(jié)合本體的有向無(wú)環(huán)圖結(jié)構(gòu),提出一種基于人類表型本體中通路的相似性計(jì)算方法。經(jīng)驗(yàn)證,該方法在預(yù)測(cè)致病基因和疾病時(shí),在不同的數(shù)據(jù)集(理想、含噪聲、含不準(zhǔn)確以及含噪聲和不準(zhǔn)確)上的效果均優(yōu)于其他基于本體的主流方法。例如本文方法在預(yù)測(cè)致病基因的含噪聲和不準(zhǔn)確數(shù)據(jù)集上比第二好的Resnik方法提高了17.3個(gè)百分點(diǎn),在預(yù)測(cè)疾病的含噪聲和不準(zhǔn)確數(shù)據(jù)集上比此方法提高了18.1個(gè)百分點(diǎn)。有研究發(fā)現(xiàn)在疾病和基因網(wǎng)絡(luò)中,同類的疾病、功能相關(guān)的基因在網(wǎng)絡(luò)中表現(xiàn)出聚集特征。病人的身上體現(xiàn)出來(lái)的表型特征中不可避免的出現(xiàn)一些和該疾病或者致病基因無(wú)關(guān)的表型(噪聲表型),利用這種聚集特性可以篩選出一個(gè)表型集中噪聲表型,可以提高致病基因和疾病預(yù)測(cè)的準(zhǔn)確率。本文構(gòu)建了表型網(wǎng)絡(luò),在表型網(wǎng)絡(luò)中利用Page Rank算法尋找中心表型和周邊表型,從而挖掘表型集中的噪聲表型,達(dá)到表型去噪的目的。經(jīng)實(shí)驗(yàn)?zāi)M,該方法能很好的發(fā)現(xiàn)噪聲表型(平均逆序數(shù)為0.136),去噪能夠提高預(yù)測(cè)致病基因和疾病的準(zhǔn)確率。
[Abstract]:The next generation of gene sequencing technology has accelerated the generation of gene data, such a large amount of data, so that today's problem from sequencing to how to effectively use the generated gene data. Phenotypic refers to the recipient gene. The phenotypes are closely related to genes and diseases. Phenotypology is now an important link between genes and diseases. Phenotypic research has been an important means to predict and find pathogenic genes. At present, there are three main research methods in gene, phenotype and disease. One is to use the existing biomedical literature. Using text mining and other related techniques to explore the relationship between biomedical entities. The other is the use of existing relationships to establish phenotypic, disease, protein, gene and other entities of the network. New relationships are found in the network. The last is the use of ontology, such as gene ontology. Phenotypic ontology and other structured knowledge systems to calculate the similarity between entities. Each of them has its own advantages and disadvantages. The study of ontology-based phenotypic similarity calculation method can help to predict the pathogenic genes and diseases of patients. Make full use of the value of ontology. This paper mainly uses human phenotypic ontology human Phenotype Ontology. HPOs are used as tools to study the similarity between genes and phenotypes and between diseases and phenotypes. Based on the phenotypic information in human phenotypic ontology, this paper combines the structure of directed acyclic graph with ontology. A method for calculating the similarity of pathways based on human phenotypic ontology is proposed. It has been proved that the method can be used in different data sets (ideal, noise-containing) in predicting pathogenic genes and diseases. Contains inaccuracies and noise and inaccuracies). For example, the proposed method is 17.3% higher than the second best Resnik method in predicting noise-containing and inaccurate data sets of pathogenic genes. The noise-and inaccurate data set for disease prediction is 18.1 percentage points higher than this method. Studies have found that similar diseases are found in disease and genetic networks. Function-related genes exhibit aggregation characteristics in the network. Some phenotypes (noise phenotypes) that are unrelated to the disease or pathogenic gene are inevitably found in the patient's phenotypic features. A phenotypic concentrated noise phenotype can be selected by using this clustering characteristic, which can improve the accuracy of the prediction of pathogenic genes and diseases. In this paper, a phenotypic network is constructed. Page Rank algorithm is used to find the central phenotype and peripheral phenotype in phenotypic network, so as to mine the noise-phenotype of phenotypic concentration and achieve the purpose of phenotypic denoising. This method can detect the noise-phenotype (the average inverse number is 0.136), which can improve the accuracy of predicting pathogenic genes and diseases.
【學(xué)位授予單位】：哈爾濱工業(yè)大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2015
【分類號(hào)】：R440;TP391.1

【參考文獻(xiàn)】

相關(guān)期刊論文 前1條

1 王秀艷;崔雷;;應(yīng)用關(guān)鍵動(dòng)詞抽取生物醫(yī)學(xué)實(shí)體間語(yǔ)義關(guān)系研究綜述[J];現(xiàn)代圖書情報(bào)技術(shù);2011年09期

，

本文編號(hào)：1457562