基于数据挖掘的乳腺癌基因筛选与药物重定位方法研究

发布时间：2018-04-10 21:31

本文选题：数据挖掘 + 基因筛选　；参考：《华东师范大学》2016年硕士论文

【摘要】：乳腺癌疾病相关基因筛选与药物研发是生物医学研究领域的重要课题,有效筛选相关转移基因和预测现有药物的新用途对于该疾病的治疗具有重要的科学意义。然而,如何挖掘和利用疾病与药物相关的特征信息一直是目前该领域的难点。随着数据挖掘技术的发展,通过特征整合与数据模型算法的结合,为以上问题提供新的解决途径。本文利用数据挖掘的特征筛选算法及分类算法研究乳腺癌相关基因筛选问题和药物重定位问题。主要研究内容如下：1.提出了一种基于随机森林分类算法和病人基因特征信息的乳腺癌转移标志基因筛选方法PPIRF。该方法的特点在于不仅考虑到基因表达值在疾病分类预测问题中的重要性,而且把蛋白质-蛋白质相互作用信息作为先验知识融合到基因挑选过程。通过与其它方法进行对比,结果表明PPIRF筛选出的基因组合具有更好的预测能力和生物可解释性。2.提出了一种基于药物多类特征信息的重排序算法Ranking-based KNN用于乳腺癌等疾病的药物重定位。Ranking-based KNN方法的特点在于不仅融合了化学结构信息、靶标信息、副作用信息以及拓扑相关性信息多种药物特征,而且通过Ranking-SVM算法得到了待测药物的最可信近邻的排序以使用加权打分策略来预测该药物的新用途。实验结果表明该方法可以帮助寻找治疗乳腺癌等类疾病的药物新用途。3.开发了名为DREP的药物重定位可视化预测工具。该工具共包含了两种药物重定位方法：第一种是基于Ranking-based KNN算法的药物重定位；第二种是基于逻辑回归分类的药物重定位。该工具可以对KEGG数据库中1387种药物与1514种疾病的未知关系进行预测,并根据预测出的药物-疾病对分值数给出所查询药物可能治疗的疾病名称列表。DREP工具界面友好,操作简单,便于生物学研究人员的使用。
[Abstract]:Screening of genes associated with breast cancer and drug research and development are important topics in biomedical research. It is of great scientific significance to screen relevant transfer genes and predict new uses of existing drugs for the treatment of breast cancer.However, how to mine and utilize the characteristic information of disease related to drugs has been a difficult problem in this field.With the development of data mining technology, the combination of feature integration and data model algorithm provides a new way to solve the above problems.In this paper, the feature screening algorithm and classification algorithm based on data mining are used to study the selection of breast cancer related genes and drug relocation.The main research contents are as follows: 1.The feature of this method is not only to consider the importance of gene expression value in disease classification prediction, but also to fuse protein-protein interaction information into gene selection process as a priori knowledge.Compared with other methods, the results show that the gene combination selected by PPIRF has better predictive ability and biointerpretability. 2.In this paper, a reordering algorithm based on multi-class characteristic information of drugs, Ranking-based KNN, is proposed for drug relocating. Ranking-based KNN method for breast cancer and other diseases is characterized by the fusion of not only the chemical structure information, but also the target information.The side-effect information and topological correlation information are used to predict the new use of the drug. Furthermore, the Ranking-SVM algorithm is used to get the ranking of the most reliable neighbor of the drug to be tested and to use the weighted scoring strategy to predict the new use of the drug.The results show that this method can help to find new use of drugs to treat breast cancer and other diseases.A visual prediction tool called DREP for drug relocalization was developed.The tool includes two drug relocalization methods: one is drug relocalization based on Ranking-based KNN algorithm, the other is drug relocalization based on logical regression classification.The tool can predict the unknown relationship between 1387 drugs and 1514 diseases in KEGG database. According to the predicted drug-disease score, the list of disease names that can be treated by the drugs queried. The DREP tool has friendly interface and simple operation.Easy to use by biological researchers.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：R737.9

【相似文献】