Dissertation > Excellent graduate degree dissertation topics show

Clustering of gene expression analysis method

Author: LiuYueMing
Tutor: YiDong
School: Third Military Medical University
Course: Medical Statistics
Keywords: Gene chip Gene expression data Clustering algorithm Self-organizing map Fuzzy Clustering Judge Entropy
CLC: R346
Type: Master's thesis
Year: 2001
Downloads: 334
Quote: 2
Read: Download Dissertation

Abstract

Objective: gene chips (the Gene chip or microarrays) is a recent molecular biology technology a breakthrough in the use of the technology can parallel analysis of the expression of thousands of genes at the same time, resulting in a mass of useful data, analyze and organize these The advantage of this technology a major bottleneck. Cluster analysis is one of the most effective means of analysis, and is the foundation of further research. Clustering of gene expression data analysis is still in its infancy, there are many issues to be resolved, one of the more prominent are the following two points: (1) algorithm and its parameters, select. Now has many algorithm applied to gene expression data analysis, the and constantly new algorithm. The clustering algorithm to select the appropriate parameters in the appropriate data to produce satisfactory results when these conditions are not met, the clustering results on the poor. Therefore, in a specific clustering problem, the main challenge is not from how to perform the clustering, but in the selection algorithm and parameter values. The current method often depends on the non-intuitive parameters, even for its statistical experts also difficult to make the right judgment to select (2) clustering results. Not an appropriate evaluation method of clustering results, the lack of an objective basis for clustering algorithm selection, the quality of clustering results also lack the necessary means of inspection. The purpose of this study is to explore these two aspects. Method: (1) fuzzy C-means method is widely used in the cluster analysis of gene expression data, but the parameter c to be artificial, through the establishment of a PFS discriminant function to solve the problem of determining parameter c, called PFS fuzzy clustering method. Clustering results judged in the FOM evaluation method based on an external evaluation standard gain ratio, the establishment of a new evaluation method-entropy evaluation method. Results: (1) the first several groups of simulated data PFS function test, to obtain satisfactory results. And PFS fuzzy clustering method to a real data clustering, compared with the data set of known functional classification, functional classification of the the PFS fuzzy clustering results with the data set showed a good correlation to verify the PFS the effectiveness of the fuzzy clustering method. ② to we create Elitropy evaluation method on the SOM method, fuzzy clustering, K-means clustering and hierarchical clustering performance in Lyer and Ferea, data sets were judged. Found that the SOM method and fuzzy clustering method clustering performance is higher than the other four clustering algorithms: K-means and average linkage algorithm and good single-chain and complete linkage algorithm. The evaluation results at the same time also to some extent validated value of Elltropy evaluation method. Conclusion: (1) the PFS fuzzy clustering is effective clustering method can be applied to gene expression data clustering analysis (2) entropy evaluation method is based on class structure and the external information of the data set in two ways to judge the clustering results, evaluation of clustering results is simple and intuitive. Entropy judge that the SOM method and fuzzy clustering method is suitable for gene expression data clustering analysis.

Clustering of gene expression analysis method

Abstract

Related Dissertations