Cancer Gene Expression Data Attribute Partial Order Representation And Knowledge Discovery

Gene chip can massively detect the expression of thousands of genes in one experiment, and it has a very important practical significance for tumor classification and diagnosis. In recent years, gene expression data show exponential growth, how to effectively analysis, organize and deal with these vast amounts of gene expression data and extract effective biological, medical information from it, which has become a hot spot of concern and research. In order to solve this problem, this article has combined feature selection method and structural partial-ordered attribute diagram(SPOAD) based on formal concept analysis(FCA) to deal with lung adenocarcinoma gene expression data, exploring a new knowledge discovery methods.In this paper, basic concept and the related definitions of formal concept analysis, structural partial-ordered attribute diagram have been researched, the advantages of structural partial-ordered attribute diagram in the study of the relationship between the knowledge discovery and visualization of data have be analyzed. And a scheme of knowledge discovery to collect new information from cancer gene expression data is proposed, which is based on the feature selection and structural partial-ordered attribute diagram. Then the paper introduces the relevant knowledge of gene expression data, the resource of lung adenocarcinoma gene expression data to be processed and pre-processing operations. Both the T-test method and the Elastic net method are used in the feature gene selection of lung adenocarcinoma gene expression data, a total of 35 feature genes have been selected, this process greatly reduces the dimension of the data.Finally, the c # program is applied to disperse the data in order to generate the form of binary formal context, then the structural partial-ordered attribute diagram is generated, and the knowledge discovery is produced based on the distribution and aggregation hierarchy diagram. Eventually, the selected genes differentially express in tumor samples and normal samples, the target genes closely related to the occurrence and metastasis of tumor have been identified, and the fact that smoking affects gene expression in tumor tissue is discovered through comparison; In addition, most of the genes have moderate expression, only a few of genes have highly expression or low expression in tumor samples.