Research On Clustering And Aligning Methods For Gene Expression Time Series Data Analysis

The mature application of high-throughput detection technology, such as cDNA microarray and oligonucleotide microarray, produced a large amount of gene expression data, including static data and time series data. The time series data reflects gene characteristics in time course. Analysis of gene expression time series data can obtain some important information, for example gene function and the relationship between genes. Currently, how to analyze the time series expression data is an important issue to be addressed in bioinformatics study.In this thesis, we studied the clustering and aligning methods for gene expression time series data analysis. The following research has been done:1. An HMM-based hierarchical clustering method for gene expression time series data was proposed, data was mapped to model space to use the time characteristics of gene expression time series data, and a hierarchical clustering strategy was used to adapt to high-throughput gene expression data clustering.2. A shortened dynamic time warping alignment method was proposed. We used a local alignment method to align gene expression time series data, and reduced computation by restricting the alignment area. The problem that aligns gene with different expression speed was solved, and the accuracy of alignment was improved.3. In order to overcome the discrete defect of gene expression time series data, a B-spline curve curvature alignment method was established. At first, used the B-spline curve to fit time series expression data, and then measured the similarity between curves by an improved curvature method.4. The proposed clustering and aligning methods were tested on specific gene expression dataset separately, such as budding yeast dataset, and the effectiveness of the proposed methods was verified.